fatigue - a complex subject - some simple approximations
TRANSCRIPT
N A S A T E C H N I C A L NASA TM X-52084 M E M O R A N D U M
s
f
x E c a v) a z
L
L > / (PAGES) I ~ O D E I I
39- ' c
: (NASA 7 r n X 5 2 d Y / CR OR TMX C H AD N U M 6 E R l ICATEOORYI
FATIGUE: A COMPLEX SUBJECT - SOME SIMPLE APPROXIMATIONS Microfiche (MF)
,
by S. S. Manson Lewis Research Center Cleveland, Ohio
WILLIAM M. MURRAY LECTURE Presented to Society for Emerimental Stress Analysis Cleveland, Ohio, October 30, 1964
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION WASHINGTON, D.C. - 1965
FATIGUE: A COMPLEX SUBJECT - SOME SIMPLE APPROXIMATIONS
by S. S. Manson
Lewis Research Center Cleveland, Ohio
WILLIAM M. MURRAY LECTURE
. Presented to the Society for Experimental Stress Analysis
Cleveland, Ohio, October 30, 1964
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
FOREWORD
It was my p r iv i l ege t o serve on the Executive Committee of t he
Society f o r Experimental S t r e s s Analysis i n 1951 when t h i s lec turesh ip
was f i r s t inaugurated. Normally our concern was l a r g e l y with f inanc ia l
matters, bu t a t t h i s meeting our subject of discussion was much more
pleasureable - how could w e be of grea te r se rv ice t o the membership of
t he Society and t o the t echn ica l community a t l a rge? One of t he members
of t he committee suggested t h a t it was about time - w e were then 8 years
old - t o i n s t i t u t e an honorary lec turesh ip within the Society. This
sounded l i k e a very good idea, and we were about t o embark on a d i s -
cussion as t o how it would be designated. Frankly my thoughts turned
i n the d i r ec t ion of some e l d e r l y hero of t h e pas t i n t h e f i e l d of s t r e s s
ana lys i s . I d id not think a t a l l along t h e l i n e s of one of t he o ther mem-
be r s who sat next t o me and said, "How about B i l l Murray?". B i l l w a s no t
i n the room a t t h e time. After t he suggestion was made, however, it
seemed l i k e the only possible idea. I wondered how I could ever have
thought i n any d i r ec t ion other than tha t . B i l l w a s a founder, he w a s
our f i rs t president , he w a s t he Executive Secretary, and our Treasurer.
Between meetings he was the e n t i r e Society and d id a l l of t he work, in-
cluding e d i t i n g of t he Journal. He d i d t h i s on donated time without
remuneration. So the suggestion not only made sense, bu t it r e a l l y w a s
t h e only r i g h t t h ing t o do. By the time B i l l came back, t he re was no
need f o r discussion. I d id not need t o be convinced and ne i the r d id
anyone e l se . The decis ion was unanimous; it was t o be the W i l l i a m M.
Murray Lecture. We asked B i l l i f he would give the f i r s t lec ture , and,
i ii
of course, he did.
been followed ever s ince. The twelve Murray l ec tu re s t h a t have been
given i n the ensuing years have a l l been a t r u e c r e d i t t o t he Society.
It i s indeed a p r iv i l ege t o take a place among those honoring B i l l , and
it i s my hope t h a t the l ec tu re t h a t I s h a l l present t o you w i l l i n small
measure, a t l e a s t , follow the high standards t h a t have already been s e t .
I n so doing he s e t a standard of excellence t h a t has
i v
. TABU OF CONTENTS
FORENARD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . SOME RECENT EXPERIMENTAL TOOLS AND TECHNIQUES . . . . . . . . . . .
1
3
PHENOMENOLOGICAL BEHAVIOR I N STRAIN CYCLING . . . . . . . . . . S t r a i n Cycling Concepts . . . . . . . . . . . . . . . . . . . Prediction of Axial Fatigue Life from Tensile Cetta . . . . .
Method of universal slopes . . . . . . . . . . . . . . . . . Comparison of Prediction Methods . . . . . . . . . . . . . . . Alternate Relation f o r Predicting Axial Fatigue L i f e i n Terms of Endurance L i m i t . . . . . . . . . . . . . . . . . . . . . .
Four-point cor re la t ion method . . . . . . . . . . . . . . .
. . 8 . . 8 . . 12 . . 1 2 . . 16 . . 1 7
. . 19
EFFECT OF YIELD STRESS AND NOTCH SENSITIVITY ON FATIGUE LIFE I N STRAIN CYCLING . . . . . . . . . . . . . . . . . . . . . . . . . 2 1
CONSIDERATION OF UNIVERSAL SLOPE FOR PLASTIC STRAIN LINE . . . . . 27
CRACK PROPAGATION AND ITS RELATION TO THX POWER LAW FOR CYCLIC LIFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
CUMULATIVEFATIGUE DAMAGE . . . . . . . . . . . . . . . . . . . . . 36 37
Aspects Relating t o Crack Propagation . . . . . . . . . . . . . . 41 Cyclic Hardening and Softening . . . . . . . . . . . . . . . . . .
APPLICATIONS INVOLVING STRAIN CYCLING DATA . . . . . . . . . . . . . 47 47 Relation Between Axial and Bending Fatigue . . . . . . . . . . . .
Rule of Thumb Relating S t r a i n Range t o Cyclic Life . . . . . . . . 53
FUNDAMENTAL ASPECTS OF FATIGUE . . . . . . . . . . . . . . . . . . . 54 Cyclic S t r a i n Hardening and Softening . . . . . . . . . . . . . . 54 Microstructural Events i n Fatigue . . . . . . . . . . . . . . . . 56 Development of Microcracks . . . . . . . . . . . . . . . . . . . . 59 Stage I1 Cracking . . . . . . . . . . . . . . . . . . . . . . . . 63 F i n a l F r a c t u r e . . . . . . . . . . . . . . . . . . . . . . . . . . 67
ACKN0WL;EDGMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
W i l l i a m M. Murray Lecture
FATIGUE: A COMPLEX SUBJECT - SOME SIMPLE APPROXIMATIONS
by S. S. Manson
L e w i s Research Center
INTRODUCTION
I have chosen as the subject of t h i s l ec tu re : Fatigue - some of i t s
This subjec t i s not a neglected one; complexities, some s impl i f ica t ions .
f i gu re 1 shows, f o r example, a p l o t of t h e number of repor t s t h a t have
appeared s ince 1950 r e l a t i n g t o fa t igue. Actually, t he l i t e r a t u r e goes
back 100 years, bu t even if we start i n 1950, we see how vas t it is.
references c i t e d here a r e se lec ted papers l i s t e d by the ASTM; it i s not
a complete compilation of a l l publications. Since t h e las t l i s t i n g was
i n 1962, I have extrapolated the curve f o r a few ensuing years based on
the " ru l e of thumb" given m e by t h e ASTM t h a t , on t h e average, t he papers
increase a t a r a t e of 15 percent per year.
i s fashioned a f t e r an approach suggested i n a p r iva t e communication
Professors Burns and Morrison of t h e Universi ty of London, and, I bel ieve,
p a r t i c u l a r l y appropriate f o r students of fa t igue. It i s qu i t e c l e a r t h a t
i f a person wished t o keep up with the l i t e r a t u r e and read one r epor t per
working day, he would f a l l behind on the order of 1 year f o r every year
t h a t he read. This would be t r u e i f he s t a r t e d with a knowledge of t h e
e x i s t i n g l i t e r a t u r e ; catching up on the backlog would be almost impossible.
In the c h a r a c t e r i s t i c fashion of fa t igue , we seem t o be approaching an
endurance l i m i t - a time a f t e r which keeping abreas t w i l l be v i r t u a l l y
impossible .
The
This method of presenta t ion
by
.
TM X-52084
2
Jus t why t h e r e a r e so many repor t s on t h i s subjec t i s indicated i n
f igu re 2. W e see the tremendous number of f a c e t s it presents , some in-
dependent of each other , some in t e r r e l a t ed , and some being merely a l t e r -
nate expressions ident i fy ing the same fac to r . Despite t he many invest iga-
t i ons that have been made, gene ra l i t i e s seem qu i t e e lus ive , and s tudies
of the bas ic f a t igue mechanism, inves t iga t ions of t h e many phenomena in-
volving engineering data, and f i n a l component t e s t i n g , w i l l continue t o
a t t r a c t the a t t e n t i o n of many inves t iga tors , and r e s u l t i n many reports .
When I undertook t o prepare t h i s lec ture , it w a s my in ten t ion t o provide
a t least a "thumb-nail" descr ip t ion of t h e s t a t e of t he a r t regarding
the b a s i c phenomenon. I had intended t o show f i r s t how complex it was,
and then proceed t o ind ica te some simple but usefu l approximations t h a t
were evolved i n recent years. I found, however, t h a t even a cursory sur-
vey would involve such lengthy discussion t h a t it would be inappropriate
f o r t he l imited time avai lable . Furthermore, so many exce l len t r e f e r -
ences have appeared i n recent years ou t l in ing our s t a t e of understanding
on the fundamental aspects of fa t igue , t h a t I would, i n the main, be r e -
peating, la rge ly with l e s s qua l i f i ca t ion , t he words of these exce l len t
authors ( r e f s . 1 t o 4 ) . My discussion w i l l l i m i t i t s e l f , therefore , t o
those fundamental aspects t h a t r e l a t e s p e c i f i c a l l y t o inves t iga t ions t h a t
have been conducted i n our laboratory, or those which e luc ida te some de-
t a i l s per t inent t o the discussion of o w r e s u l t s . For t h i s reason the
r e s u l t s of our experiments w i l l be described f i rs t , and the fundamental
aspects reserved u n t i l t he terminal sec t ion .
Before proceeding, however, t o a ' d i scuss ion of experimental r e s u l t s
and t h e i r s ignif icance, it i s appropriate t o mention several of t he new
3
t oo l s t h a t have r ecen t ly been arrayed toward obtaining an understanding
of the f a t igue phenomenon. This i s appropriate because t h i s Society is,
a f t e r a l l , devoted t o experimental matters.
SOME RECENT EXPERIMENTAL TOOLS AND TECHNIQUES /
Since/ f a t igue i s s o in t imate ly r e l a t ed t o the question of generat ing L -
and propagating a crack, and s ince the photoe las t ic technique' r e a l l y
comes i n t o i t s own when studying s t r e s s d i s t r i b u t i o n i n the v i c i n i t y of
a crack, it i s very l o g i c a l t h a t t h i s research t o o l should have contr ib-
uted t o our understanding of fa t igue. Figure 3 shows some r e s u l t s t h a t
have been presented by Gerberich ( r e f . 5) showing the s t r a i n d i s t r i b u -
t i o n a t the t i p of a notch. The photoelast ic coat ing method i s espe-
c i a l l y valuable here because it enables us t o s tudy the s t r a i n d i s t r i b u -
t i o n i n meta l l ic mater ia ls .
i The phase in te r fe rence method has a l s o been appl ied t o study s t r e s s -
d i s t r i b u t i o n i n the v i c i n i t y of a notch t i p . Figure 4 i s taken from an
SESA publ icat ion i n which Oppel and H i l l ( r e f . 6) have presented t h e i r
r e s u l t s using t h i s technique. Brief ly , l i g h t i s d i rec ted through an
o p t i c a l system onto an i n i t i a l l y f l a t surface of t h e specimen. I n the
undeformed condition of the specimen, t he p l a t e g l a s s i s p a r a l l e l t o t he
specimen surface. Deformation of the specimen by loading causes a va r i a -
t i o n i n the d is tance between the p la te g l a s s and the specimen a t d i f f e r -
en t po in ts , and r e s u l t s i n a pat tern very similar t o Newton's r ings , such
as t h a t shown on t h e r i g h t i n t h e figure.
Figure 5 shows another technique t h a t has been used i n our own
laboratory. It makes use of a polycarbonate r e s i n , a p l a s t i c material,
which, when fat igued, c l e a r l y shows s t r i a t i o n s . Each of these s t r i a t i o n s
4
provides an ind ica t ion of t he dis tance t h a t the crack has progressed
from one cycle t o the next. Most s p e c i f i c a l l y our i n t e r e s t i s i n how
t h e fa t igue crack ge t s s t a r t e d and how it grows during t h e very e a r l y
stages.
t h e progression of cracks without a microscope.
t o some r e s u l t s obtained i n our laboratory making use of t h i s technique.
Figure 6 r e f e r s t o another very valuable t o o l t h a t has been adopted
The use of t h i s mater ia l permits t h e visual observations of
We s h a l l r e f e r l a t e r
by inves t iga tors i n f a t igue within the l as t decade.
e lec t ron microscope, whereby very high magnifications can be accom-
pl ished, making it possible t o observe the c lose ly spaced ind iv idua l
s t r i a t i o n s during f a t igue crack growth of long- l i fe specimens. This
f igu re shows the s t r i a t i o n s t h a t have been observed by Bedesem and
Har re l l of our laboratory i n a study of t he aluminum a l l o y 7075-T6.
Every r ipp le represents the growth of the crack during a cycle of
loading. I n t h i s case t h e crack per cycle i s of t h e order of 30 micro-
inches.
r ep l i ca t ing mater ia l , as shown i n the upper l e f t of t he f igure . The
p l a s t i c material i s then s t r ipped off t he specimen, leaving an exact
r e p l i c a of t he surface topography. This surface i s then shadowed by
t h e use of carbon, or other mater ia ls such as germanium, which forms
an extremely t h i n f i l m having the shape of t h e surface.
an angle produces h ighl ights t h a t m a k e it poss ib le t o d i s t ingu i sh
regions of surface depression from those of e leva t ion .
r e p l i c a is then removed and t h e t h i n carbon or meta l l ic f i l m i s s tudied
i n the electron microscope.
This t o o l i s the -
Such pa t te rns are made by covering t h e surface with a p l a s t i c
Shadowing a t
The p l a s t i c
Figure 7 shows another approach involving t ransmission e l ec t ron
.
.
5
microscopy ( r e f . 7 ) .
t he surface, i s placed i n the microscope. To be penetrated by the e l ec t ron
beam, however, the specimen must be very th in . The l aye r s of atoms a c t as
a d i f f r a c t i o n grat ing, and any d isar ray i n the l a t t i c e i s indicated by t h e
pa t t e rn generated by t h e e l ec t ron beam on the screen.
i s shown i n f igu re 8 taken from the work of Grosskreutz ( r e f . 8) . The
dark regions a r e d is loca t ions - regions of atomic d i sa r r ay - and it i s
seen from the photographs how the grains break up i n t o a number of smaller
regions ca l led subgrains, separated from each other by d i s loca t ion ne t -
works. Since t h i s method involves the use of very t h i n f i l m s of t he
specimen i t s e l f , it i s sometimes questioned whether t he behavior of such
f i l m s i s t y p i c a l of t he behavior of bulk material . Much needs t o be done
t o resolve t h e questions t h a t have been r a i sed regarding t h e appl ica-
b i l i t y of the r e s u l t s obtained by these invest igat ions. It i s c lear ,
however, t h a t t h i s technique can be extremely useful.
A combination of t h e technique involving the use of t h e e l ec t ron
Here the specimen i t s e l f , r a t h e r than a r e p l i c a of
A t y p i c a l pa t t e rn
microscope and the Moir6 method i s i l l u s t r a t e d i n f igu re 9 ( r e f . 9 ) .
a very t h i n l aye r of a s ingle c r y s t a l of palladium was deposited on a
correspondingly t h i n s ing le c r y s t a l of gold. The a r r a y of atoms i n each
of these c r y s t a l s cons t i tu ted the grids t h a t combined t o form a Moirg
pa t t e rn . The in te r fe rence pa t t e rn produced by the gr ids w a s observed
with an e l ec t ron microscope. I n t h i s f igure , one of t h e g r ids i s assumed
t o be per fec t and the other t o contain a dis locat ion.
t h a t r e s u l t s i s d is t inguishable by means of t he e l ec t ron microscope,
whereas the a c t u a l atomic a r r a y of the mater ia ls themselves i s t o o f i n e
t o be resolved i n t h i s manner. Thus, by t h i s method, it i s possible t o
Here
The Moir6 pa t t e rn
,
6
observe disarrays i n the atomic s t r u c t u r e such as the edge d i s loca t ion
shown a t the r i g h t of t he f igure.
X-rays are an a l t e r n a t e means of de tec t ing d i sa r r ay i n a c r y s t a l
l a t t i c e , s ince the l a t t i c e w i l l d i f f r a c t t he X-rays i n much the same
manner as an e lec t ron beam. Thus, using very t h i n f i l m s of c r y s t a l l i n e
mater ia ls i n which the l a t t i c e arrangements a r e not per fec t and bom-
barding these with an X-ray beam, r e s u l t i n photographs similar t o those
shown i n f igu re 10 ( r e f . 10).
a r r a y of d i s loca t ions i n s i l i c o n , as brought out by t h e use of X-ray
d i f f r ac t ion .
This f igu re shows a very complicated
Decoration and etching techniques have been used t o provide photo-
graphic evidence of t he exis tence of d i s loca t ions and t h e i r geometric
propert ies . Figure 11 shows a c l a s s i c photograph obtained by Dash
( r e f . 11) of a Frank-Read source using copper decoration.
ment of e tch p i t s i s now a wel l -establ ished technique. Recently, a
very i n t e r e s t i n g innovation has been developed by Hahn and Rosenfeld
( r e f . 1 2 ) ; by using a s i l i c o n s t e e l and a s p e c i a l e tchant , they were
ab le t o br ing out i n bold con t r a s t t he regions of p l a s t i c flow i n t h e
v i c i n i t y of a notch. Figure 1 2 shows some of t h e i r r e s u l t s . A shee t
with a notch w a s subjected t o loading, and the region of p l a s t i c flow
w a s f i r s t observed by t h i s technique a t t h e surface. By machining away
ha l f of the specimen and then e tch ing t h e midsection the d i s t r i b u t i o n
of p l a s t i c flow i n the midsection of t he p l a t e w a s made apparent.
the difference i n appearance of these regions. Such s tud ie s can be
extremely usefu l i n understanding the r o l e of p l a s t i c flow and t h e e f f e c t
of cons t ra in t of such p l a s t i c flow on the generat ion of cracks.
The develop-
Note
7
Another technique t h a t has been qu i t e u se fu l i n studying the cracks
generated by f a t igue i s ca l led "taper sect ioning," and it i s i l l u s t r a t e d
i n f igu re 13. Here the attempt i s made t o observe a very shallow crack
by tak ing advantage of the geometric f a c t t h a t , i f t he surface i s cu t by
a plane t h a t makes a very small angle with it, the shallow crack w i l l be
elongated i n accordance with the cosecant r e l a t i o n of t he angle between
the c u t t i n g plane and the surface. Although the f igure shows t h i s
angle 8 t o be of appreciable magnitude, so as t o be geometrically
d is t inguishable , t he angle i s made very small i n p rac t i ce so t h a t t he
magnification achieved i s qu i t e large, of t h e order of 20.
cracks that a r e hardly dis t inguishable become evident when t a p e r sect ion-
i ng i s applied. It i s evident t h a t a d i s t o r t i o n of t h e geometry r e s u l t s
because the magnification occurs only i n t h e depth of crack, bu t not i n
the width; thus, care must be exercised i n in t e rp re t ing the r e s u l t s of
such s tudies . A t ape r sec t ion o f a brass specimen t e s t e d i n to r s ion i s
shown on the r i g h t of f i gu re 13 ( r e f . 13).
In t h i s way,
Another approach t h a t can be used i n the study of crack propagation
i s shown i n f igu re 14; it makes use of u l t r a son ic techniques and has been
under inves t iga t ion i n our laboratory by K l i m a , Lesco, and Freche. By
t h i s technique, it has been possible t o de t ec t cracks l e s s than 0.003 inch
long i n notched shee t specimens. As a matter of i n t e r e s t , it should be
noted t h a t cracks were detected wel l within the f i rs t 10 percent of t he
l i f e of a notched specimen. The question of crack i n i t i a t i o n i n smooth
specimens w i l l be taken up l a t e r .
The foregoing techniques are ce r t a in ly not t he only ones t h a t have
been brought t o bear on t h i s problem of fa t igue . Among o thers a r e
8
acous t ic emission, exo-electron emission, vacuum s tudies , magnetic
e f f e c t s , res i s tance e f f e c t s , damping, heat conduction, thermographic
s tud ie s , op t i ca l microscopes, cinematography, and many others. By far,
t h e most powerful t o o l i s the process of human reasoning.
last 15 years many new concepts have been generated, and t h e evaluat ion
of these concepts has produced immense s t r i d e s i n t h e understanding of
t h e fa t igue phenomenon and t h e appl ica t ion of laboratory da ta toward
cont ro l of f a t igue f a i l u r e .
Within t h e
PHE3TOMENOLOGICAL BEHAVIOR I N STRAIN CYCLING
Although the techniques j u s t out l ined f o r studying t h e fundamental
mechanism of fa t igue have produced some very use fu l r e s u l t s leading t o
a b e t t e r understanding of t h e mechanism, we are as ye t far from being
ab le t o apply our fundamental understanding t o solve the f a t igue problem
i n the f i e ld . P rac t i ca l approaches s t i l l involve t h e d i r e c t measurement
of fa t igue proper t ies of mater ia ls . Some of these r e s u l t s may be use fu l
as a f i r s t s t e p i n design and may a l s o provide some ins igh t i n t o the
f a t igue mechanism.
been obtained i n OUT laboratory ( r e f s . 1 4 t o 15).
S t r a i n Cycling Concepts
We s h a l l now discuss some recent r e s u l t s t h a t have
Consider a specimen t h a t i s subjected t o axial reversed s t r a i n cycling.
F i r s t a pos i t ive s t r a i n (+E) i s appl ied t o the specimen; by revers ing the
load, a negative s t r a i n ( - E ) i s obtained r e s u l t i n g i n a t o t a l s t r a i n range
of Z E . ~ Note t h a t t h i s type of loading i s s l i g h t l y d i f f e r e n t from what
has conventionally been done i n f a t igue t e s t i n g wherein t h e load is t h e
primary var iable t h a t i s control led.
The f i r s t observation made when a specimen i s s t r a ined i n t h i s
.
9
manner is that it requires different loads to accomplish a desired amount
of strain, depending on the number of prior applications of the strain.
This fact is illustrated in figure 15. Here, for reference purposes, a
continuous curve is plotted to illustrate static stress-strain behavior
of the material. The stress is plotted, however, as a range, indicating
the total range in stress including both the compression and the tension
halves of the cycle. Some manipulation is required to transform a con-
ventional stress-strain curve to one involving stress range and strain
range, but.it is essentially the stress-strain curve of the material.
In the first cycle of loading, the stress range required to produce a
fixed strain range, for example 0.018, is at point A in figure 15(a). It
lies on the static stress-strain curve, because, by definition, the static
stress-strain curve provides the stress required to produce a given strain
in the first quarter cycle of loading. However, if an attempt is made to
apply the same strain in successive cycles, it takes a greater stress
range to maintain the strain range. This fact is illustrated in fig-
ure 15(b) along the line A'PA. In this case, after about 600 cycles of
loading, however, the stress stabilizes and for the remainder of the test,
out to about 1400 cycles, the stress range remains approximately constant.
This achievement of an asymptotic stress range is sometimes referred to
as "saturation hardening."
I
-
If the asymptotic stress range that is reached
is plotted against the applied strain range, this point falls on
figure 15(a).
A" in
In a similar manner, if the strain range is raised to 0.036,
the variation B'% is obtained, resulting in the point B" in fig-
ure 15(a) as the asymptotic stress range associated with the strain
range 0.036. The curve joining A" and B" represents what is known as
the "cyclic stress-strain curve." In this case it lies considerably above
10
t he s t a t i c s t r e s s - s t r a i n curve; t h i s material i s thus r e fe r r ed t o as a
cyc l i c strain-hardening mater ia l .
i n f igure 1 5 ( c ) , which i s merely a r e p l o t of f i gu re 15(b) with t h e cyc l i c
l i f e scale made l i nea r . It i s apparent from f igu re 15(c) t h a t s a tu ra t ion
hardening i s achieved during the e a r l y cycles of loading, usua l ly wel l
before half the number of cycles t o failure have been applied.
Figure 1 6 i l l u s t r a t e s t h e behavior of another c l a s s of mater ia l
O f add i t iona l i n t e r e s t i s t h e p l o t
described as cyc l i c s t r a i n softening. I n t h i s case, f o r a s t r a i n range
of 0.015, t h e s t r e s s range required i s i n i t i a l l y a t poin t
quickly diminishes u n t i l , a f t e r a r e l a t i v e l y f e w cycles, t he s t r e s s
range i s only about ha l f t h a t a t point A ' . The cyc l i c s t r e s s - s t r a i n
curve here i s
curve. The degree of hardening or sof ten ing va r i e s considerably among
mater ia ls , and some mater ia ls even show hardening and sof ten ing i n t h e
same t e s t .
A, bu t it
AI'B'' and fa l ls considerably below t h e s t a t i c s t r e s s - s t r a i n
Before present ing an ana lys i s of t h e data, it i s important t o d i s -
cuss some d e t a i l s of procedure used i n our inves t iga t ions . Some of t h e
conf l ic t ing repor t s t h a t have appeared i n t h e l i t e r a t u r e regarding
s t ra in-cyc l ing data a r e due i n part t o a lack of c l e a r descr ip t ion of
t he procedure t h a t i s followed i n obtaining t h e data.
A s a l ready indicated, the t e s t s r e fe r r ed t o here (ref. 15) a r e for
s t r a i n cycling about zero mean s t r a i n .
t ransverse s t r a i n range (diametral) r a t h e r than t h e a x i a l s t r a i n range
w a s maintained constant.
of t r u e s t r a i n i n the region when failure occurs.
g l a s s shape w a s found most convenient f o r t hese tests.
I n most of these t e s t s t h e
This w a s done i n order t o produce known values
A specimen of hour-
When analyzing
.
L
11
t he data, the l i f e was taken as t h e number of cycles required t o cause
complete rupture of the specimen.
here because some inves t iga tors regard the l i f e of a specimen as the
point a t which cracks begin t o appear on the surface.
ence between these two concepts w i l l be in te rpre ted .
It i s important t o make a d i s t i n c t i o n
Later t h e d i f f e r -
When sa tu ra t ion hardening o r sof tening a c t u a l l y occurs, t he value of
In some s t r e s s range i n the cyc l ic s t r e s s - s t r a i n curve i s unambiguous.
cases, where hardening or sof ten ing continues throughout t h e t e s t , t h e
value of s t r e s s range used i s t h a t a t t he h a l f - l i f e of the specimen.
When a t o t a l s t r a i n range A€ ( = 2 ~ ) i s applied t o a specimen ( i n
our t e s t s t he a x i a l s t r a i n range i s deduced from a knowledge of s t r e s s
range, diametral s t r a i n range, and the e l a s t i c cons tan ts ) , t h a t part
of the t o t a l s t r a i n range, which i s e l a s t i c , can be separated from t h a t
which i s p l a s t i c by a knowledge of the s t r e s s range.
range i s simply t h e s t r e s s range divided by the e l a s t i c modulus. The
The e l a s t i c s t r a i n
p l a s t i c s t r a i n range i s obtained b y subt rac t ing t h e e l a s t i c s t r a i n range
from the t o t a l s t r a i n range. Figure 1 7 shows idea l ized r e s u l t s t h a t are
obtained when fa t igue l i f e i s p lo t ted aga ins t e i t h e r t he e l a s t i c or
p l a s t i c s t r a i n range.
r i thmic.
of t hese two components, i s a l s o shown i n the f igure .
S t r a igh t l i nes r e s u l t when both sca l e s a r e loga-
The curve depict ing t h e t o t a l s t r a i n range, which i s t h e sum
The concept t h a t cyc l i c l i f e i s r e l a t e d t o t h e p l a s t i c s t r a i n range
w a s f i r s t proposed by us i n 1952 ( r e f . 16 ) . This r e su l t ed from our
e f f o r t t o es t imate the importance of temperature on the thermal s t r e s s
f a t i g u e of tu rb ine buckets.
have been made by many inves t iga tors t o ve r i fy the r e l a t ion .
Since then, a considerable number of s tud ie s
The l i n e a r -
L
12
i t y of e l a s t i c range with l i f e has been proposed more r ecen t ly ( r e f . 17)
i n an e f f o r t t o overcome the d i f f i c u l t y , from a p r a c t i c a l point of view,
of determining l i f e from t h e p l a s t i c range only.
s t r a i n that i s known r a t h e r than the p l a s t i c s t r a i n , and the re fo re it i s
d i f f i c u l t t o es t imate how much of t h i s s t r a i n i s p l a s t i c unti l t he
e l a s t i c component i s known.
i n pract ice , t o r e l a t e l i f e t o the t o t a l s t r a i n , which i s the sum of t h e
e l a s t i c and p l a s t i c components.
It i s usua l ly the t o t a l
For t h i s reason, it i s much more useful ,
It w i l l be noted t h a t t he t o t a l s t r a i n range i s not a s t r a i g h t l i n e .
Because of the logarithmic sca l e involved, t he curve of t o t a l s t r a i n
range against l i f e i s asymptotic t o t h e p l a s t i c l i n e i n t h e lower cyc l ic -
l i f e range, and t o the e l a s t i c l i n e i n t h e higher c y c l i c - l i f e range.
Predict ion of Axial Fatigue Li fe from Tensi le Data
From a p r a c t i c a l point of view, merely knowing t h a t t he e l a s t i c and
p l a s t i c components a r e approximately s t r a i g h t l i n e s i s extremely use fu l
because it means t h a t only a few t e s t s a r e needed t o e s t a b l i s h these
l i n e s .
avoid any f a t igue t e s t i n g whatsoever and t o es t imate these s t r a i g h t l i n e s
from a knowledge of more r e a d i l y obtained mater ia l p roper t ies .
fore , an attempt has been made t o e s t a b l i s h a co r re l a t ion between these
l i n e s and t h e proper t ies of mater ia l s obtained i n simple t e n s i l e t e s t s .
There a re , however, many appl ica t ions when it i s des i rab le t o
There-
Four-point-correlation method. - Figure 18 shows one method of anal-
y s i s tha t w e have invest igated. It may be r e fe r r ed t o as a four-point
method because the two s t r a i g h t l i n e s a r e obtained by loca t ing two poin ts
on each of them.
t e n s i l e behavior of t he mater ia l . A po in t i s located on t h e e l a s t i c l i n e
Every point i s determined from a knowledge of the
13
a t 1/4 cycle with an ordinate ( 2 . 5 ur)/E, where of i s the t r u e f r a c t u r e
s t r e s s of t he mater ia l obtained by dividing the load a t the time of f a i l -
ure i n the t e n s i l e t e s t by the ac tua l a r ea measured a f t e r f a i l u r e has
occurred.
point , the ordinate i s (0.9 uu)/E, where
t e n s i l e s t rength of the material .
10 cycles i s determined t h a t has an ordinate of
logarithmic d u c t i l i t y of t he mater ia l defined as the n a t u r a l logarithm
of the o r i g i n a l c ross -sec t iona l a rea of the specimen divided by t h e f i n a l
c ross -sec t iona l area. The second point on the p l a s t i c l i n e is obtained
a t lo4 cycles as indicated i n f igu re 18.
lo4 cycles i s f i r s t located on the e l a s t i c l i n e and the ordinate observed.
This ordinate i s then subs t i t u t ed in to the simple equation shown i n t h e
f igure t o obtain a corresponding ordinate value a t lo4 cycles f o r t h e
p l a s t i c s t ra in . This formula i s derived from the observation t h a t t he
p l a s t i c and the e l a s t i c s t r a i n s a t lo4 cycles a r e approximately r e l a t e d
t o each other. The r e l a t i o n i s almost (but not qu i te , see r e f . 15) equiva-
l e n t t o t h e assumption t h a t t he t o t a l s t r a i n range a t lo4 cycles i s
approximately 1 percent f o r a l l materials. Thus, from a knowledge of t he
t e n s i l e proper t ies , two points on each of t he l i n e s can be determined
and t h e p l a s t i c and e l a s t i c components p lo t t ed as i n f igu re 18. The curve
of t o t a l s t r a i n range may be obtained by simple addi t ion, as indicated by
the curved l i n e , which i s asymptotic t o the p l a s t i c l i n e i n the low-l i fe
range and t o the e l a s t i c l i n e i n the h igh - l i f e range. This curved l i n e
thus represents t h e estimated r e l a t ion between t o t a l s t r a i n range and
l i f e f o r t h e mater ia l .
Another point on t h i s l i n e i s obtained a t lo5 cycles. A t t h i s
uu i s the conventional u l t imate
On t h e p l a s t i c l i ne , a poin t a t
D3/4 , where D i s the
The poin t shown by t h e star a t
14
I n order t o obtain t h e fa t igue proper t ies i n t h i s way, t he t r u e
f r a c t u r e s t r e s s must be known; however, t h i s property i s not always
given i n the l i t e r a t u r e , and therefore an add i t iona l approximation i s
sometimes required.
M r . John O'Brien of San Diego, Cal i fornia , who, under cont rac t with
NASA, recommended t h a t t he f r ac tu re s t r e s s could be obtained by multi-
plying the ul t imate t e n s i l e s t rength by the f ac to r
A very good approximation has been suggested by
1 + D. Thus,
u f = u u (1 + D) (1)
That t h i s r e l a t i o n i s v a l i d i s seen i n f igu re 19, where f r a c t u r e s t r e s s
i s p lo t ted aga ins t t he product of u u ( l + D ) .
s en t s a d i f f e ren t mater ia l , and the data general ly f a l l c lose t o a
45O l i ne .
two t e n s i l e proper t ies , uu
a r e needed t o pred ic t a x i a l f a t igue l i f e f o r a spec i f ied s t r a i n range.
This i s i l l u s t r a t e d i n f igu re 20. One simply loca te s the value of
and the value of percent reduction i n a rea on the hor izonta l ax is , and
then determines the in t e rcep t s a t P1 and P4 from the dashed and
s o l i d famil ies of curves. The loca t ion of points P2 and Pg are ob-
ta ined from the aux i l i a ry v e r t i c a l and hor izonta l s ca l e s i n the f igure .
Each data point repre-
By using t h i s approximation, only the e l a s t i c modulus and
and reduction i n a rea (which e s t ab l i shes D)
uu/E
The procedure described for determining t h e e l a s t i c and p l a s t i c
l i n e s was f i r s t developed by s tudy of r e l a t i v e l y few mater ia ls .
sequently, the v a l i d i t y of t h i s procedure w a s inves t iga ted by examining
a l a r g e r number of mater ia ls .
a x i a l low cycle f a t igue t e s t s t o da te a r e shown i n t a b l e I. Alloys of
near ly a l l the important c l a s ses of s t r u c t u r a l mater ia l s a r e included.
Sub-
Those mater ia l s t h a t have been s tudied i n '.
15
TABLJI I. - MATERIALS FOR AXIAL LOW FATIGUE INVESTIGATION
4130 Sof t
4130 Hard
4130 X-hard
4340 Annealed
4340 Hard
304 Annealed
304 Hard
52100 Hard
52100 X-hard
AM 350 Annealed
AM 350 Hard
310 S ta in l e s s
Vascomax 300 CVM
Yascojet MA
Vascojet 1000
~
Titanium 6A1-4V
Titanium 5A1-2.5Sn
Magnesium AZ3lB-F
Aluminum 1100
Aluminum 5456 H 3 1 1
Aluminum 2014 T6
Aluminum 2024 T4
Aluminum 7075 T6
S i l v e r 0.99995 pure
Beryllium
Inconel X
A286 Aged
A286 34 Percent cold reduced and aged
D9 79
These mater ia l s cover qu i t e a range i n va r i ab le s t h a t might a f f e c t
Among them a r e crys- f a t i g u e behavior such as those shown i n t a b l e 11.
t a l l i n e s t r u c t u r e s wherein body-centered-cubic, face-centered-cubic, and
hexagonal-close-packed arrangements a re represented; reductions i n a rea
covering the range from 1 t o 94 percent; t e n s i l e s t r eng ths from 16 000 t o
over 400 000 pounds per square inch; high and low notch s e n s i t i v i t i e s ;
c y c l i c hardening and sof ten ing cha rac t e r i s t i c s ; high and low s tacking
f a u l t energy; e t c . Thus, any conclusion t h a t might be reached regarding
16
TABLE 11. - MATERIAL VARIABLES I N AXIAL L O W CYCLE
FATIGUE INTTESTIGATION
Crys ta l l ine s t r u c t u r e
Methods of s t rengthening
Reduction i n a rea
Tensile s t r eng th
True f r a c t u r e stress
E l a s t i c modulus
Notch s e n s i t i v i t y
Stacking f a u l t energy
Cyc 1 i c be havi or
Body centered cubic
Face centered cubic
Close packed hexagonal
P rec ip i t a t ion hardening
Hot and cold worked
1 To 94 percent
16 000 To 411 000 p s i
48 000 t o 500 000 p s i
6 . 2 ~ 1 0 ~ To 4 2 . 0 ~ 1 0 ~ p s i
Notch d u c t i l e t o very notch s e n s i t i v e
Low ( s tee ls ) t o high (aluminum)
3 t r a in sof ten ing t o s t r a i n hardening
the v a l i d i t y of t h e r e l a t i o n s involving a l l of these materials cannot be
regarded as 'being l imi ted t o a very small c l a s s of materials.
Method of un ive r sa l slopes. - An a l t e r n a t e approach i s t o assume t h a t I
t he slopes of t h e e l a s t i c and p l a s t i c l i n e s are t h e same f o r a l l materials.
In reference 1 7 t h e universa l s lope f o r t h e p l a s t i c l i n e was assumed t o
be -0.5 and for t h e e l a s t i c l i n e -0.12. Data obtained more r ecen t ly have
permitted t h e se l ec t ion of more ref ined values, as indicated i n f i g -
ures 2 1 and 22. Here t h e r e s u l t s of tes ts on 29 materials are shown.
For t h e p l a s t i c l i n e ( f i g . 2 1 ) a l l t h e data can be represented reason-
ab ly w e l l by a s t r a i g h t l i n e with a slope of -0.6. For t h e e l a s t i c l i n e
a l l t he data can be represented reasonably w e l l by a s ing le s t r a i g h t
,
17
l i n e of slope -0.12. The l i n e s of f igures 2 1 and 22 a l s o permit t h e
determination of a s i n g l e point through which the l i n e s of the known
slopes may be drawn. Thus, l e t t i n g %= 1 i n each of the equations
i n t h e two f igu res shows the in te rcept of t he p l a s t i c l i n e a t I!@= 1
i s
construct ion of t he l i n e s of universal s lope i s shown i n f igu re 23,
where it i s ind ica ted t h a t , s t a r t i n g a t I@= 1 and ordinates of D O g 6
N.6, and the in t e rcep t of the e l a s t i c l i n e i s 3.5 au/E. The bas i c
3.5 au and ., E , respect ively, f o r the p l a s t i c and e l a s t i c l i nes , these
l i n e s a r e then completed by drawing slopes of -0.6 and -0.12, respec-
t i ve ly . The equation f o r t o t a l s t r a i n range AE then becomes
where
a, ul t imate t e n s i l e s t rength, p s i
D d u c t i l i t y , In 1 1 - R A
RA reduct ion i n a rea , percent
Nf number of cycles t o f a i l u r e
Figure 24 provides an add i t iona l a i d t o the construct ion of the e l a s t i c
and p l a s t i c l i n e s based on equation ( 2 ) using two points on each l i n e
(a t 10 and a t lo4 cyc les ) . The i d e n t i f i c a t i o n of points P1 t o P4'
and t h e determination of t he ordinates from a knowledge of only
and a,/E may be obtained from the f igure . The arrows on each curve
ind ica t e t h e appropriate sca les t o be used.
FiA
Comparison of Prediction Methods
Figures 25 t o 27 present the ac tua l comparisons between t h e experi-
ments and the pred ic t ions for each of t h e 29 mater ia ls t e s t e d when ana-
18
lyzed on the bas i s of each of t h e methods indicated. The ind iv idua l da t a
poin ts for t o t a l s t r a i n range ( i . e . , t h e sum of t h e e l a s t i c and p l a s t i c
components) are shown as c i r c l e s . The predic t ions by t h e four-point
co r re l a t ion ( inc luding t h e knowledge of the f r a c t u r e s t r e s s ) a r e shown
by t h e dashed l i n e s .
slopes a re shown by the s o l i d l i n e s .
remarkably good.
t h e agreement i s not as sa t i s f ac to ry .
Predictions based on the method of un ive r sa l
I n most cases t h e agreement i s
For AM 350 annealed ( f i g . 26), and beryll ium ( f i g . 2 7 ) ,
An ove ra l l evaluation of t h e four-point method is shown i n f i g -
ure 28. Here t h e measured l i f e f o r each of t he tes t poin ts i s p lo t t ed
aga ins t the predicted l i f e from a knowledge of t h e f r a c t u r e s t r e s s ,
u l t imate t e n s i l e s t rength , and d u c t i l i t y . The r e l a t i o n between t h e
da t a points and t h e l i n e i s ind ica ted by t h e t a b l e a t t h e lower r i g h t .
Thirty-five percent of t h e da t a points f a l l within a f a c t o r or 1.5 i n
l i f e from t h e predicted value.
f a c t o r of 5. Since some s c a t t e r i n l i f e i s expected i n f a t i g u e da ta ,
it can be seen that t h i s co r re l a t ion must be regarded as s a t i s f a c t o r y .
A few data points, notably those assoc ia ted with poorly behaved mate-
r ials l i k e beryllium, a r e f a i r l y remote from t h e co r re l a t ion l i n e ,
thus r e su l t i ng i n t h e conclusion ind ica ted by t h e t a b l e t h a t , with an
allowable e r r o r of a f a c t o r of 20 i n l i f e , only 97 percent of t h e da ta
points w i l l be sa t i s f ac to ry .
t h e b a s i s of measured and predicted s t r a i n range r a t h e r than on t h e
b a s i s of l i f e . The t o t a l s t r a i n range predicted t o produce t h e l i f e
observed in each t e s t i s p lo t t ed on t h e ordinate.
s t r a i n range i s p lo t t ed on t h e abscissa. The co r re l a t ion i s , as might
Almost 90 percent f a l l within a l i f e
S imi la r ly , f i gu re 29 compares t h e d a t a on
The measured t o t a l
19
be expected, considerably b e t t e r . Ninety percent of t he da ta poin ts
f a l l within a s t r a i n f ac to r of 1.5.
Figure 30 shows a corresponding ove ra l l comparison on a l i f e b a s i s
f o r t h e method of un iversa l slopes. Similar ly , f i gu re 31 shows the
comparison on the b a s i s of predicted s t r a i n f o r t he same 29 materials
as obtained by the method of universal slopes.
experiment i s , i n both cases, approximately the same as t h a t obtained
by the four-point method.
The agreement with
Alternate Relation f o r Predict ing Axial Fatigue
Li fe i n Terms of Endurance L i m i t
I n t h e preceding discussions, s t r a i n has been expressed i n terms
of the e l a s t i c and p l a s t i c components, both of which were l i n e a r when
p lo t t ed aga ins t l i f e on log-log coordinates. This approximation i s
extremely use fu l i n p r a c t i c a l appl icat ions.
express l i f e d i r e c t l y i n terms of s t r a i n , however, t he r e s u l t i n g
equation does not r e a d i l y lend i t s e l f t o a closed-form expression.
There i s a l s o another minor objection t o the manner of r e l a t i n g s t r a i n
and l i f e thus far discussed. This i s the f a c t t h a t it does not provide
f o r t h e p o s s i b i l i t y of the existence of an endurance l i m i t s t r a i n , a
s t r a i n below which the l i f e i s e s s e n t i a l l y i n f i n i t e . Figure 32 i nd i -
ca t e s an a l t e r n a t e r e l a t ion , which, i n a sense, overcomes both of
these objections. Cyclic l i f e Nf i s expressed i n terms of t he appl ied
t o t a l s t r a i n range
p l a s t i c . This equation i s
When one attempts t o
A€, without regard t o what percentage i s e l a s t i c o r
Nf = A(Ac - DE^)^ ( 3 )
Included i n the equation i s a constant, Ace, which i s an endurance l i m i t
20
s t r a i n . Below Ago, l i f e i s taken as i n f i n i t e . This expression w i l l
obviously coincide with the p l a s t i c s t r a i n equation when AE i s large,
as i s the case i n the very low-cycl ic- l i fe range. That is , f o r la rge
values of A€ the value of Ago becomes negl ig ib le , and a power l a w
r e l a t i o n between l i f e and p l a s t i c s t r a i n resul ts , Thus, t he term A
would be expected t o be of the order of the coe f f i c i en t i n the p l a s t i c
s t r a i n r e l a t i o n and v t o be of t he order of t he exponent of t h e
p l a s t i c s t r a i n r e l a t ion .
For the purpose of determining whether a r e l a t i o n of t h e type shown
i n f igu re 32 i s va l id f o r t he 29 mater ia l s , an attempt was made t o de t e r -
mine the three constants A, Ago, and v by s a t i s f y i n g the equation a t
th ree points : cyc l i c l i v e s of 10, 1000, and 100 000 cycles. One ap-
proach involved the se l ec t ion of t h ree poin ts on the experimental curve
determined from ac tua l data.
se lec ted from t h e curve predicted by t h e method of un iversa l slopes and
the constants subsequently obtained.
i n f igu re 32.
experimental data and the curve of un iversa l s lopes has been emphasized
for the sake of c l a r i t y . Figures 33 t o 35 show the a c t u a l comparison
between the experimental and predicted curves f o r a l l 29 mater ia ls by
using equation (3).
In a second approach, t h ree points were
Both these methods a r e i l l u s t r a t e d
The degree of discrepancy shown i n t h i s f i gu re between t h e
Since the s t r a i n values a t the lives indica ted by t h e stars on the
dot-dash curve ( f i g . 32 ) can be expressed e x p l i c i t l y i n terms of duc-
t i l i t y and ul t imate s t rength , it i s apparent t h a t A, Ae0, and v can
a l s o be determined e x p l i c i t l y i n terms of t hese two simple measurable
propert ies . The a lgebra ic r e l a t i o n s t h a t r e s u l t , however, are qu i t e
. 21
complicated and w i l l not be described here.
made i s i l l u s t r a t e d i n f igu res 33 t o 35.
l i n e s i l l u s t r a t e t he agreement between the da ta and the curves obtained
by means of the equation shown a t the t o p of t he f igu res when the con-
s tan ts a r e determined from a knowledge of t h e experimental da ta a t th ree
The important po in t t o be
I n these f igu res t h e s o l i d
values of l i f e .
t h e constants were determined from curves drawn by t h e method of uni-
v e r s a l slopes. Obviously, corresponding curves could be drawn t o repre-
sen t t he points as predicted by the four-point approximation method. I n
each case, the fit coincides almost exac t ly with t h e corresponding f i t
associated with the method of universal slopes, and i s therefore not
shown.
l i f e i n terms of s t r a i n range can be represented by an equation of t h e
type shown i n equation (3) and t h a t , where appropriate , t h i s equation
may be used ins tead of t h e ones discussed previously.
The dash-dot curves provide a similar comparison when
The ma in feab re t h a t i s demonstrated i s t h a t t he equation f o r
EFFECT OF YIELD STRESS AND NOTCH SENSITIVITY
ON FATIGUE LIFE IN STFAIN CYCLING
The previous discussion emphasized the importance of ul t imate ten-
s i l e s t r eng th and reduct ion i n area as p r inc ipa l var iab les governing
s t r a i n cycl ing f a t igue behavior. Before drawing any f i n a l conclusions,
however, it is important t o consider t h e poss ib le s ign i f icance of other
proper t ies - f o r example, y ie ld s t ress and notch s e n s i t i v i t y . That
y i e l d s t r e s s cannot be a property of primary inf luence on the f a t igue
c h a r a c t e r i s t i c s can be seen from f igures 15 and 16, which have a l ready
been discussed i n connection with the cyc l i c s t r e s s - s t r a i n r e l a t ion .
Since only a very few cycles can d r a s t i c a l l y change the s t r e s s required
t o produce -a
i s of l i t t l e
22
given s t r a i n , it can be seen t h a t t he i n i t i a l y i e ld stress
meaning i n t h e major number of cycles within a given f a t igue
test.
dismissed as a property of primary importance i n governing fa t igue l i f e
i n the low-cycle range.
The yield s t r e s s as determined i n a v i rg in mater ia l can thus be
The second var iable , notch s e n s i t i v i t y , requi res more extensive
consideration before i t s proper r o l e can be assessed.
vary appreciably i n t h e i r notch cha rac t e r i s t i c s , and t h a t notch sens i -
t i v i t y i s c l a s s i c a l l y linked t o f a t igue r e s i s t ance i s wel l es tabl ished.
Since t h i s property has not entered i n t o t h e co r re l a t ion thus far d i s -
cussed, it i s des i rab le t o question the appropriateness of t he omission,
and i f appropriate, then why?
That materials
I n order t o a sce r t a in whether t h e f a t igue r e s u l t s a r e a f fec ted by
notch sens i t i v i ty , a series of tests w a s conducted i n which the same
mater ia l w a s heat t r e a t e d t o achieve d i f f e r e n t degrees of notch sens i -
t i v i t y while maintaining t h e same t e n s i l e s t r eng th and d u c t i l i t y . The
b a s i s of t h e t e s t s i s indicated i n f igu re 36. Here published da ta f o r
s t a i n l e s s s t e e l 410 a r e shown. The center curve shows t e n s i l e s t r eng th
as a function of tempering temperature.
a t e i the r 400° or 850° F produces approximately t h e same t e n s i l e s t rength.
Correspondingly, t h e reductions i n area f o r these two heat treatments
a r e almost ident ica l .
curve, however, i s very se r ious ly a f f ec t ed by the tempering temperature.
Tempering a t 850' F produces a mater ia l of much lower impact res i s tance
than the one tempered a t 400' F.
t h e 850' F temper produces a notch s e n s i t i v e material .
It can be seen t h a t tempering
The impact res i s tance as indicated by t h e lowest
From t h i s it might be infer red t h a t
We have con-
2 3
ducted notch tests on sheet a t our laboratory with t h i s s t e e l t h a t ve r i -
f i e d t h i s conclusion.
As a r e s u l t , w e had the opportunity of t e s t i n g two mater ia l s with i d e n t i c a l
t e n s i l e s t rengths and d u c t i l i t y b u t far d i f f e r e n t notch s e n s i t i v i t y values.
The 850° F temper w a s extremely notch sens i t ive .
"he a x i a l s t r a i n cycl ing fa t igue 'behavior of t h i s material i n these
two conditions of temper i s shown i n f i gu re 37.
t o t he 400' F temper and closed data poin ts r e f e r t o t h e 850° F temper
condition.
t h a t t h e two mater ia ls produced approximately the same fa t igue r e s i s t -
ance; t he re was qui te a d i f fe rence i n the appearance of t he f r a c t u r e
surface, however.
b r i t t l e f r a c t u r e surfaces, while i n many cases the 400' F tempered mate-
r i a l showed very d u c t i l e f r a c t u r e surfaces. Nevertheless, as far as
c y c l i c l i f e i s concerned, the two r e s u l t s were qu i t e similar.
Open data poin ts r e f e r
Although the re is a small amount of s c a t t e r , it can be seen
The 850° F tempered material f requent ly showed very
I n order t o understand these r e su l t s , it i s necessary t o undertake con-
s ide ra t ion of crack occurrence i n a f a t igue t e s t of the kind t h a t i s being
considered here.
i n i t i a t i o n of a crack by severa l methods. One of t he more i n t e r e s t i n g
approaches has been by the use of a polycarbonate r e s i n mater ia l . Ad-
mit tedly, because t h i s mater ia l i s a p l a s t i c , it does not
i n with the f a t igue ana lys i s t h a t has been presented.
has been qu i t e usefu l i n gaining an understanding of crack growth.
s h a l l f i rs t present t h e data obtained with t h i s mater ia l because it has
a v i s u a l a t t r ac t iveness . A t the same time it should be recognized t h a t
t h e conclusions drawn w i l l not be inconsis tent with the behavior of
me ta l l i c mater ia ls .
I n our laboratory, we have attempted t o observe t h e
qu i t e f i t
Nevertheless, it
We
Figure 5 has already been used t o show why poly-
24
carbonate mater ia ls a r e s o usefu l i n the study of crack growth. S t r i a -
t i o n s a r e c l e a r l y evident on the f r a c t u r e surface on a macroscopic sca le .
Each cycle i s c l e a r l y del ineated and can r e a d i l y be determined. The par-
t i c u l a r specimen of f igu re 5 had a l i f e of 115 cycles.
from t h e las t cycle, it i s possible t o determine exac t ly where crack
growth stopped a t each cycle of loading.
of t h e data fo r t h i s specimen and f o r one which had a l i f e of 1300 cycles.
The crack depth i s p lo t ted aga ins t percent of f a t igue l i fe . The l i m i t of
reso lu t ion of these cracks w a s of t h e order of a depth of 0.002 t o 0.003
inch.
order of t h i s magnitude starts qu i t e l a t e i n the l i f e of t h e specimen,
approximately 65 percent of t h e l i f e f o r t h e low-cycle t e s t and 85 per-
cent f o r t he high-cycle t e s t .
s u l t s of a number of t e s t s conducted t o determine approximately when,
i n the l i f e of an a x i a l l y cycled specimen, the crack reaches a s i z e of
0.002 t o 0.003 inch. Several types of data a re shown i n t h i s f igure .
The da ta represented by open c i r c l e s a r e taken from Laird and Smith
( r e f . 18).
mens of pure aluminum and nickel.
polycarbonate data already discussed.
obtained a t our laboratory with 2024-T4 aluminum from surface observa-
t i ons . The da ta correspond t o the percent of t o t a l l i f e required t o
produce surface crack lengths of the order of 0.010 inch, which were
v i s i b l e without magnification.
and 410 s t a i n l e s s s t e e l . These data were a l s o obtained from surface
observations.
By counting back
Figure 38 shows the ana lys i s
It i s c l e a r from t h i s f igure , however, t h a t a crack depth of t h e
Figure 39 provides a summary of t he r e -
They obtained t h e i r r e s u l t s by counting s t r i a t i o n s on speci-
The closed c i r c l e s represent t he
The t r i a n g l e s represent da ta
Also shown a r e r e s u l t s f o r 4130 s t e e l
For l a t e r use, it w i l l be des i reable t o express t h e
25
da ta i n f igu re 39 by an a n a l y t i c a l expression.
aga ins t Nf
by t h e equation
A p l o t of 1 - (No/Nf)
on log-log coordinates r e s u l t s i n a s t r a i g h t l i n e defined
1 - (No/Nf) = 2.5 NP1I3 (4)
where
No number of cycles t o v i s i b l e cracking
Nf number of cycles t o complete f r ac tu re
This equation i s obviously va l id only for values of Nf higher
than 15 cycles. For values of Nf below 15 cycles, t he equations would
imply t h a t N,/Nf i s negative. Equation ( 4 ) i s p lo t t ed i n f igu re 39
and i s seen t o f i t the data qu i t e w e l l , although data are , of course,
inadequate t o ensure t h a t a l l materials can be represented by a s ing le
curve. An important point t o be observed from t h i s curve i s t h a t only
f o r t h e very low-cycl ic- l i fe t e s t s does a crack of any appreciable s i z e
occur e a r l y i n the l i f e of a specimen. For cyc l i c l i v e s f o r t he order
of 1000 or more, de tec tab le cracking does not occur u n t i l near ly 70 per-
cent or more of t he cyc l i c l i f e has been used up.
Now it is axiomatic t h a t a m a t e r i a l , which is s e n s i t i v e t o notches
or cracks, w i l l not manifest i t s s e n s i t i v i t y u n t i l the crack a c t u a l l y
develops. I f t h e crack does not occur u n t i l qu i t e l a t e i n the l i f e of
t h e specimen, t h e mater ia l cannot be s e n s i t i v e t o it u n t i l most of t he
l i f e has already been used up. Since the type of t e s t under considera-
t i o n here, namely, a small specimen subJected t o a x i a l cyc l ic s t r a i n
loading, does not involve cracking until very l a t e i n i t s l i f e , it might
be expected t h a t the behavior of the mater ia l i n t h e presence of a crack
w i l l no t s i g n i f i c a n t l y a f f e c t the r e su l t . Even i f the material were t o
. 26
f a i l immediately on the development of a small crack, the l i f e would be
reduced on the order of 50 percent a t a l i f e of 100 cycles , and much
l e s s f o r the higher cyc l i c l ives .
reduction might be d i f f i c u l t t o de t ec t i n the ana lys i s of t he data.
i s concluded, therefore , t h a t t he important f ea tu re of a x i a l s t r a i n
cycl ing t e s t s i s the very l a t e development of s i g n i f i c a n t cracking and,
therefore , t h a t t he p r inc ipa l reason why t e s t s of t h i s kind do not re -
f l e c t notch s e n s i t i v i t y i s t h a t cracks simply a r e not present i n these
mater ia ls during the major port ion of t h e i r l i f e .
Considering normal da ta s c a t t e r , t h i s
It
One might ask whether t,hese t e s t s a r e the re fo re r e a l i s t i c s ince
many component p a r t s a r e indeed operated i n the presence of cracks.
Furthermore, i n many cases, t he re a r e l a rge bodies i n which crack
propagation cons t i t u t e s a grea t port ion of t he l i f e .
w i l i be touched on later i n t h e sec t ion on cumulative f a t igue damage.
A t t h i s point, i t i s important t o recognize t h a t t he t a sk we have set
f o r ourselves w a s t o es t imate the ax ia l f a t i g u e l i f e of a p a r t i c u l a r
specimen in a p a r t t c u l a r type of t e s t , one t h a t i s f requent ly used,
and the one t h a t w a s used t o obtain the da ta shown i n the previous
f igures .
r e s u l t s of these a x i a l f a t igue t e s t s from t h e t e n s i l e proper t ies of
t he materials and whether notch s e n s i t i v i t y should be considered a
primary variable. What we have shown i s t h a t t h e answer i s negative
for the type of specimen employed i n the t e s t s , and t h a t t he r e s u l t s
of these t e s t s can b e s t be estimated by using only t h e t e n s i l e s t r eng th
and d u c t i l i t y measured on smooth specimens.
meaningful i n terms of component operat ion i s a quest ion equal ly t o be
This question
!!%e problem we s e t about t o discuss w a s how t o estimate t h e
Whether t h e r e s u l t s a r e
27
asked of any da ta obtained on small, nonnotched specimens, e i t h e r by
experiment o r by est imat ion by the procedures indicated. The f a c t t h a t
t h e est imates of f a t igue l i f e of small, smooth specimens must be care-
f u l l y in t e rp re t ed before they can be appl ied t o es t imate t h e l i f e of a
s t ruc tu re i s an important consideration t h a t w i l l be discussed l a t e r i n
conjunction with cumulative f a t i g u e damage.
CONSIDERATION OF UNIVERSAL SLOPE FOR PLASTIC STRAIN LINE
While we a r e on the subjec t of when, during the course of t h e
l i f e of a f a t igue specimen, detectable cracking f i r s t occurs, it i s
appropriate t o r e v e r t t o the i n t e r e s t i n g question of t h e magnitude of
t he s lope of t he p l a s t i c s t r a i n l i ne . When we f i r s t proposed i n 1952
t h a t t h e p l a s t i c s t r a i n l i n e i s l i n e a r on log-log coordinates, t he pro-
posal w a s based on very l imi ted experimental data ava i l ab le a t t h a t
time.
of a universa l slope f o r a l l materials.
or slope, would be a mater ia l constant. I n l a t e r inves t iga t ions Coffin
( r e f . 19) suggested t h a t t h e exponent has a universa l value of -0.5 f o r
a l l mater ia ls , and f o r many materials inves t iga ted s ince then, he has
concluded t h e value of -0.5 t o be the appl icable exponent. On the other
hand, t he t e s t s t h a t have been conducted i n our s tudies of many mater ia ls
i nd ica t e t h a t t h e slopes a r e indeed d i f f e r e n t from mater ia l t o material,
bu t t h a t i f a universa l slope i s t o be assumed, then a value of -0.6
would be more representa t ive of a l l mater ia ls . It i s important, there-
fore , t o reconci le the various observations by d i f f e r e n t inves t iga tors .
Before f u r t h e r discussion of t h i s subjec t can be undertaken, it i s
Certainly the re were not enough data t o j u s t i f y t h e assumption
We suggested t h a t the exponent,
important t o consider what i s meant by t h e f a i l u r e l i f e of a specimen. I n
some of h i s t e s t s ,
cracking occurred,
may a l s o have used
28
Coffin observed the specimen surface, and when v i s u a l
he regarded the t e s t complete. Other i nves t iga to r s
i n i t i a l cracking as the c r i t e r i o n f o r f a i l u r e . I n a l l
of t he t e s t s on t h e 29 mater ia ls i n our program, f a i l u r e was taken as
the ac tua l separat ion of t h e two halves of t he specimen. We can now,
therefore , examine the r e l a t i o n between these two c r i t e r i a f o r f a i l u r e
by t h e use of f igu re 40. This f igu re i l l u s t r a t e s the r e l a t i o n between
p l a s t i c s t r a i n range and l i f e f o r an idea l ized mater ia l i n which the two
cases of specimen separat ion and of cracking a r e considered.
t inuous l i n e shown i n t h i s f i gu re is an idea l ized l i n e , t y p i c a l of
severa l of t he materials inves t iga ted showing a slope of -0.6. Since
t h i s l i n e presumably cons t i t u t e s specimen separation, and s ince we have
already indicated in f igu re 39 a r e l a t i o n between cracking and separa-
t i o n , we may now apply the curve of f igu re 39 t o es t imate when i n i t i a l
cracking might have occurred for t h i s idea l ized material .
s e l ec t ing seve ra l l i f e values on t h e s o l i d l i n e , it may be determined
from f igure 39 when i n i t i a l cracking probably occurred.
a r e r ep lo t t ed as the c i r c l e s shown i n f i g u r e 40, It can be seen t h a t ,
when one passes an approximate s t r a i g h t l i n e through these c i r c l e s , it
has a slope of -0.5.
l i n e can be extrapolated back t o an in t e rcep t a t 1/4 cycle , which i s
lower than the continuous l i n e . A discrepancy i n in t e rcep t has a l s o
been noted i n the l i t e r a t u r e when cracking i s considered as ind ica t ive
of cyc l ic l i f e compared with specimen separat ion. It i s not, of course,
c e r t a i n t h a t t h i s i s the complete explanation f o r t h e discrepancy. Such
a wide var ia t ion i n slopes e x i s t s among various mater ia l s t h a t many
The con-
Thus, by
These l i v e s
A l s o of i n t e r e s t i s t h e f a c t t h a t t h i s s t r a i g h t
29
other fac tors probably contr ibute t o it. Nevertheless , t h i s idea l ized
ana lys i s i s i n t e r e s t i n g and provides some ins igh t as t o the possible
reason f o r discrepancies i n published r e s u l t s .
CRACK PROPAGATION AND ITS RELATION TO POWER LAW
FOR CYCLIC LIFE
Many attempts have been made t o explain the power l a w r e l a t i o n be-
tween p l a s t i c s t r a i n range ( o r s t r e s s range) and l i f e , b u t no completely
s a t i s f a c t o r y r a t i o n a l explanation has as y e t been offered.
ana lys i s may be made by examining the r e l a t i o n f o r crack growth.
following discussion an attempt w i l l be made t o provide a very approxi-
mate r e l a t ion by using a crack growth concept.
t h i s approach is not a t a l l rigorous b u t may serve as a s t a r t i n g point
from which a more va l id ana lys i s may be made.
some of the f ea tu res required of an ana lys i s of t h i s type i n order t o be
consis tent with experimental observations.
One type of
I n t h e
It i s recognized t h a t
A t l e a s t it points t o
We s t a r t with an analogy t o observations of crack growth i n s t r e s s
cycling. It has been observed t h a t t h e r a t e of crack growth per cycle
i n s t r e s s cycl ing i s a power l a w of s t r e s s and crack length. Studies of
crack growth i n reversed s t r a i n cycl ing a r e very l imited, bu t i n general
we may observe from the l imi ted information ava i l ab le t h a t t he parameter
(where i s t h e p l a s t i c s t r a i n range and 1 i s the crack
length) is about as fundamental i n s t r a i n cycl ing as
cycling.
and
A a d i s i n s t r e s s
Thus, we assume a power l a w r e l a t i o n between crack growth r a t e
30
where s is a constant as ye t t o be determined. If the crack growth
rate i s d i r e c t l y proport ional t o Aep o r t o $, then s = 1, which
w i l l l a t e r be seen t o be a des i r ab le assumption i n order t o s a t i s f y t h e
quan t i t a t ive observations of t h e p l a s t i c power l a w exponent.
From equation (5)
d Z = C(AcP)' dN * In t eg ra t ing equation ( 6 ) between ( Z0,N0) and ( 2 f , N f ) , where 2, and No
a r e any corresponding values of 2 and N during which the crack growth
l a w , equation ( 6 ) , i s va l id , and where
cyc l i c l i f e a t f r a c t u r e Nf , r e s u l t s i n
2f i s t h e crack length a t the
2 - s 2 -s - 2 kF - ZT) = C(Acp)s(Nf - No) = C(AcP)'Nf(1 - 2) ( 7 ) 2 - s
The f i r s t problem i s t o s e l e c t values of 2, and No. A s already
indicated, these could be any corresponding values of 2 and N during
which equation ( 6 ) is val id . Frequently, (e.g., ref. 20) t h i s d e t a i l i s
overcome by taking No t o be small i n comparison with Nf, whi le 2,
i s small compared with Z f ; thus, such an ana lys i s proceeds by omitt ing
both t h e 2, and No terms i n equation ( 7 ) , r e s u l t i n g i n t h e r e l a t i o n
-
By f u r t h e r assuming t h a t t h e crack length a t f r a c t u r e i s approximately
constant , t he p l a s t i c power l a w r e l a t ion follows immediately from equa-
t i o n (8) . I n f a c t , i f s i s taken as 2, w e ge t the universa l slope of
-1/2 on a p l o t of l og Aep
the in t eg ra t ion of equation ( 6 ) is n o t va l id as shown s ince it requi res
aga ins t log Nf, although f o r a value of s = 2
31
a logarithmic i n t e g r a l of d l / l .
f ea tu res of the argument, and t h e same conclusion i s reached by t r e a t i n g
the equation from the very beginning f o r
T h i s does not change the e s s e n t i a l
s = 2 and in t eg ra t ing properly.
While the preceding approach leads t o a power l a w r e l a t i o n as sought,
t he re are some unsa t i s fy ing fea tures associated with it. F i r s t , although
any set of corresponding values of lo and No can be chosen as con-
s t a n t s of in tegra t ion , it must be ensured t h a t these values apply within
t h e range i n which t h e crack growth l a w (eq. ( 5 ) ) i s val id . The question
then develops i n t o one of es tab l i sh ing whether t h i s r e l a t i o n i s indeed
v a l i d i n t h e very e a r l y s tages of cycling when the crack i s too small
t o be detected and i s perhaps confined t o ind iv idua l grains within an
i n i t i a l l y smooth specimen. For, i f the crack growth equation i s not
v a l i d i n these e a r l y s tages , No cannot be taken as negl ig ib le i n com-
par ison with Nf wi thin the range of v a l i d i t y of equation (5 ) .
To avoid t h i s complication, a refinement i s introduced whereby we
choose, instead, t h e values of No and lo corresponding t o the l i f e
a t which r i p p l e s or s t r i a t i o n s a r e f i r s t formed on the f r a c t u r e surface,
or a t which v i s i b l e surface cracking can be observed. Then we can be
reasonably c e r t a i n t h a t the specimen i s i n the crack growth s tage f o r
which equation (5) w a s assumed t o be va l id , while a t the same time not
denying t h a t equation (5) could possibly have been v a l i d a t an e a r l i e r
s tage of t h e l i f e . Choosing such a small value of lo permits us t o
omit t h e
2f term (provided t h a t s < 2; i f s > 2 it might, i n f a c t , be per-
missible t o omit t he 2f term). We cannot, however, neglect t h e No
term on the r i g h t s i d e of the equation. Since we have a l ready deter-
-
lo term on the l e f t s ide of equation ( 7 ) , compared with t h e
32
mined, i n equation (4), t he value of 1 - (No/Nf) as 2.5 Nf1l3, we may
subs t i t u t e t h i s value i n equation ( 7 ) , leaving us with t h e expression
2.5C(Acp) Nf s 2 / 3 on the r i g h t s ide of equation ( 7 ) .
A second refinement i s t o consider t he crack length a t f r ac tu re .
I n t h e consideration of b r i t t l e mater ia ls , the G r i f f i t h equation
constant has been amply ve r i f i ed . How t h i s equation i s t o be modified
f o r a duc t i l e mater ia l i n s t r a i n cycl ing has not been es tab l i shed , bu t
a f i r s t approach i s as follows: Using the conventional s t ra in-hardening
expression fbr the cyc l ic s t r e s s - s t r a i n curve
MJZ' =
t h e assumption of an expression similar t o the G r i f f i t h equation would
r e s u l t i n
AU& = A U , ( A E ~ ) ~ & = B (10)
I n order t o allow f o r some d i f fe rence between t h e f r a c t u r e l a w i n the
e l a s t i c range and t h a t i n t h e p l a s t i c range, we w i l l assume t h a t t he
r e l a t i o n i n t h e p l a s t i c range i s (Ao)q& = const
constant ye t t o be determined.
where q
Thus, equation (10) becomes
i s a
( A U ) q G = ( A U , ) ~ ( A ~ , ) ~ ~ * = B
Thus, subs t i t u t ing f o r lf from equation (11) i n t o equation
neglecting lo, and applying equation (4), y ie ld
2 [ B ( A U ~ ) - ~ ( A ~ ~ ) - ~ ~ ] ~ - ~ = 2 .5 C(Acp) s Nf 2/3 2 - s
Combining terms involving AeP, w e a r r i v e a t
33
s < 2 2
31s + qn(2 - s ) ] where w =
and where
s power l a w exponent for crack growth l a w ; 1 i f crack growth r a t e i s
proport ional t o A E ~ or ,fi
n s t r a i n hardening exponent of cycl ic s t r e s s - s t r a i n curve; between
approximately 0.1 and 0.3
q modification f a c t o r fo r p l a s t i c f r ac tu re derived from G r i f f i t h
equation f o r e l a s t i c f r ac tu re ; 1 if same r e l a t i o n i s followed
p l a s t i c a l l y as e l a s t i c a l l y
A p l o t of w aga ins t nq f o r severa l values of s i s shown i n
f igu re 41. It can be seen t h a t , fo r n = 0.2 ( an average value obtained
from experimental da ta ) and q = 1 (assuming an equation of the G r i f f i t h
form t o be v a l i d ) , a value of s z 1 i s required i n order t o make
0.5. It can a l s o b e seen t h a t high values of s, f o r instance, above
1.5, would r e s u l t i n values of w well below the experimentally ob-
served range, whereas values of s = 1.0 would produce very reasonable
values. Thus, i f t h i s theory i s correct , the implicat ion i s t h a t , i n
t h e crack growth r a t e l a w (eq. (5)), the crack growth r a t e i s more near ly
proport ional t o nep+ than it i s t o ( A G , ) ~ ~ .
w =
It has already been observed tha t t he case s = 2 cannot be t r e a t e d
i n t h e manner j u s t described, because of t h e logarithmic r e l a t i o n t h a t
develops during the in t eg ra t ion of equation ( 6 ) .
however, be r e a d i l y ca r r i ed out, and computations show t h a t t he value
of w
by f i g u r e 41 f o r values of s t h a t approach 2.0.
The procedure can,
i s very close t o 1/3 f o r a l l values of nq, as would be implied
34
The preceding ana lys i s introduces an i n t e r e s t i n g question regarding
the crack growth exponent. If the ana lys i s i s va l id , it would appear
t h a t a value of s i n the v i c i n i t y of 1.0 would b e s t explain t h e r e l a -
t i o n between p l a s t i c s t r a i n and cyc l i c l i f e . Experimental determina-
t i o n s would seem t o point , however, t o values of s = 2 as the more
reasonable assumption, as noted, f o r example, i n f igu re 38(b) where the
logarithm of crack depth i s p lo t ted aga ins t percent l i f e t o obtain very
near ly l i nea r r e l a t ions . The implication, then, i s t h a t t he in tegra-
t i o n should involve a term log 2 , as would be obtained i f s were
equal t o 2. The question revolves, however, about t he meaning of t he
parameter, p l a s t i c s t r a i n range, when t h e crack i s growing deep enough
t o be measurable. In our t e s t s , we measure t h e diametral s t r a i n and
maintain it constant even a f t e r the crack starts; i n the last s tages
of t he t e s t we maintain a x i a l displacement constant. Thus, t he re i s
r e a l l y no uniform p l a s t i c s t r a i n across the sec t ion when the crack i s
of appreciable measurable depth. If w e consider only t h e l imi ted region
of 2 , f o r instance, l e s s than 0.01, we have t o question whether l i n e a r -
i t y i n t h i s range on a logarithmic sca le ( i n f i g . 38(b) ) has any r e a l
s ignif icance, and whether equl ly good l i n e a r i t y would not be obtained
on other sca les associated w i t h other values of s. Thus, w e must con-
clude t h a t t he i n d i r e c t method of determining s f o r t h i s geometry by
examining t h e exponent w might have some merit over the method of
d i r e c t measurement. Since the re a r e so many unanswered questions regard-
ing t h e v a l i d i t y of a l l ava i l ab le analyses of t h e problem, it must be
concluded tha t considerably more research i s needed t o c l a r i f y them.
An a l t e rna te approach t h a t leads t o acceptable values of t h e
35
exponent i n the power l a w r e l a t i o n is t o separate the exponents of
and 2 i n equation (5) . For example, we could assume t h a t
cP
d l t s / 2 - = C(AEP) 1 dN
and proceed i n the same manner as before t o in t eg ra t e t h i s d i f f e r e n t i a l
equation. The f i n a l r e s u l t , analogous t o equation (13) i s
AE Nw = const P
where
2 3[t + qn(2 - s ) ]
w =
On t h i s bas i s , t he re a r e a la rge number of admissible combinations of s
and t t h a t produce values of w i n t h e desired range of 2 / 3 . Specif-
i c a l l y , f o r example, the s implest combination i s f o r s = 2, t = 1. For
t hese values, a logarithmic in tegra t ion i s required, r e s u l t i n g i n t h e r e -
la t i on
2 f
20 I n - = 2 0 5 C A E ~ N ~ / ~
The introduct ion of a f r ac tu re r e l a t i o n f o r of analogous t o
equation (10) destroys the exact l i n e a r i t y of t he logarithmic p lo t of
Acp aga ins t N f J bu t the curvature involved i s r e l a t i v e l y small. In
f a c t , t h e s implest procedure f o r i n t e rp re t ing equation (15) i s t o
recognize t h a t , f o r t h i s combination of values of s and t, the curve
of crack growth aga ins t number of cycles increases very r ap id ly i n the
region of f a i l u r e . Thus, r a the r than introducing a f a i l u r e c r i t e r i o n
based on s t r e s s or AcPJ we can say t h a t f a i l u r e w i l l occur a t the same
value of 2f f o r a l l values of AcP; because of the s t e e p slope of the
2 - N curve, l a rge d i f fe rences between the t rue and assumed values of
lf w i l l have a small e f f e c t on Nf. With lf and lo having f ixed
36
values of a given mater ia l fo r a l l values of
e s s e n t i a l l y t h a t A E ~ N ~ ~ / ~ = const.
AcP, equation (15) s t a t e s
The purpose of t he foregoing der iva t ions is , of course, not t o o f f e r
an accurate der iva t ion of t he power l a w r e l a t ion , bu t r a t h e r t o ind ica te
some of the component considerations t h a t must be taken i n t o account i n
such a der ivat ion i f it i s based on the process of crack growth. The
most important f a c t o r i s the establishment of an accurate crack growth
l a w (preferab ly on a physical basis, bu t here a r b i t r a r i l y hypothesized
f o r s impl ic i ty) ; accounting f o r t he nucleat ion period (i. e., not simply
neglect ing
some bas is f o r f r a c t u r e ( i . e . , a c r i t e r i o n f o r choosing
as a l l of these s teps involve power l a w r e l a t ions , t he f i n a l equation i s
i n the form of a power l a w r e l a t i o n between hep and Nf. By i n t r o -
ducing equation (9), a corresponding power l a w r e l a t i o n develops between
s t r e s s range and cyc l i c l i f e . Thus, t he model hypothesized i n figure 17,
r e l a t i n g both e l a s t i c and p l a s t i c components t o cyc l ic l i f e becomes
val idated on t h e b a s i s of these power l a w r e l a t ions .
CUMULATIVE FATIGUE DAMAGE
The t e s t s described s o far r e l a t e t o the behavior of mater ia l i n
No, bu t somehow accounting f o r it); and f i n a l l y e s t ab l i sh ing
2 f ) . A s long
which the independent var iab le , s t r a i n range, i s maintained constant
throughout t he l i f e of the specimen.
a c t u a l s t ruc tures , we must t ake i n t o account operat ion under a spectrum
of loading. This i s normally r e fe r r ed t o as cumulative fa t igue . The
c l a s s i c a l l i n e a r damage ru l e f i rs t propcsed by F’almgren ( r e f . 21) and
l a t e r again by Langer ( r e f . 22) and by Miner ( r e f . 23) i s now w e l l known.
It assumes t h a t a t any s tage of t he loading h i s t o r y of t he mater ia l , t h e
When considering the behavior of
.
37
percentage of l i f e used up i s proportional t o t h e cycle r a t i o a t t h a t
loading condition. Thus, i f a s t r e s s range or s t r a i n range i s appl ied
f o r , n l cycles a t a condition where failure would occur i f N1 cycles
a r e applied, t h e percentage of l i f e used up i s That t h i s i s j u s t
an approximation and may often r e su l t i n erroneous predict ions of f a t i g u e
l i f e i s w e l l known.
the f a c t o r s t h a t govern mater ia l behavior under cumulative f a t igue load-
ing and see wherein t h i s r u l e tends t o f a l l down, and t o determine t h e
requirements of a ru l e t h a t would be more accurate.
nl/N1.
Perhaps it would be appropriate t o out l ine some of
Cyclic Hardening and Softening
Let us consider f i rs t the behavior of a mater ia l under cumulative
f a t i g u e loading as manifested by observation by i t s cyc l i c hardening
and sof tening cha rac t e r i s t i c s . Figure 42 shows some r e s u l t s obtained
with 2024-T4 aluminum under an interrupted loading spectrum.
curve AB represents t he load range p lo t t ed aga ins t cycles f o r a
t e s t i n which the s t r a i n range is maintained constant a t 0.070, and f o r
which the l i f e i s 27 cycles. The curve CD represents a corresponding
behavior when the s t r a i n range is maintained a t 0.021 and the l i f e i s
657 cycles.
of 0.070 i s appl ied f o r only 2 5 percent of t he expected l i f e f o r t h i s
s t r a i n l eve l , it w i l l b e noted t h a t , during t h i s period, t he va r i a t ion
of load range follows the curve AE. If a t t h i s t i m e t he s t r a i n range
i s changed t o 0.021 and t h i s l a t t e r s t r a i n range i s maintained u n t i l
f a i l u r e occurs, t he r e s u l t i n g load range curve i s E 'F . When examined
i n t h e type of p l o t shown i n f igu re 42(a) , t he curve
what complicated because, during the l a t t e r 75 percent of the t e s t when
The
Now i f the t es t i s conducted s o t h a t t he s t r a i n range
E ' F appears some-
38
t he s t r a i n range i s 0.021, the mater ia l f i r s t sof tens and then hardens.
When viewed on t h e type of p l o t such as f igu re 42(b) , however, t h i s be-
havior becomes more understandable. Here, t h e hor izonta l s ca l e i s taken
t o be percent of l i f e r a t h e r than cyc l i c l i f e , otherwise the curves i n
both p lo ts represent t he same phenomena.
hardening from A t o E follows the bas i c curve AB, as would be ex-
pected.
t o f a l l within a r e l a t i v e l y few cycles t o t h e curve
t o t h e same percentage of l i f e t h a t was already used up on curve AB.
Thus, the curve E'F f a l l s rap id ly a t f irst t o curve CD, bu t s ince a t
25 percent of t he expected l i f e t he curve
s t i l l undergoing considerable hardening, t h e curve E'F proceeds i n a
manner t o ind ica te f u r t h e r hardening.
It i s seen t h a t , i n i t i a l l y the
When t h e strain range changes t o 0,021, t he stress range tends
CD t h a t corresponds
CD ind ica t e s the material i s
A similar r e s u l t i s obtained i f one starts f i rs t with t h e low s t r a i n
range and then continues t h e t e s t a t the higher s t r a i n leve l .
t h e curves separated we have chosen t o i l l u s t r a t e t he case i n which ha l f
of t h e l i f e i s used up a t t h e lower s t r a i n l eve l , followed by s t r a i n i n g
a t t h e higher leve l . The f i r s t port ion of t h e t es t i s represented by CG,
which f a i t h f u l l y follows CD; the second follows the curve G'H. I n
f i g u r e 42(a), t h i s segment i s disconnnected from t h e o ther curves and
seems t o bear l i t t l e assoc ia t ion t o them.
l i f e , however, ( f i g . 4Z(b)), the s ign i f icance of G'H i s made c lear . It
r i s e s rapidly t o the v i c i n i t y of curve AB, and follows it c lose ly there-
a f t e r , indicat ing the v a l i d i t y of a l i n e a r cumulative l i f e r u l e when con-
s ide r ing the degree of hardening achieved a t any poin t i n the h i s t o r y of
loading.
To keep
When p lo t t ed aga ins t percent
39
More extensive t e s t s are shown i n f igu re 43 f o r the same material .
I n f igu re 43(a) , the r e s u l t s a r e shown f o r a change of s t r a i n l e v e l a t
approximately 25 percent of l i f e for both a high- and a low-strain- level
test . I n f igu re 43(b), t h e change is made a t approximately 50 percent
of l i f e and i n f igure 43(c) two changes a r e introduced - one a t approxi-
mately 25 percent and the other a t approximately 50 percent of l i f e . It
is seen t h a t i n a l l cases t h e tendency i s almost ident ica l .
a f t e r the s t r a i n range i s changed, the curve tends t o seek a s t r e s l e v e l
Very s h o r t l y
corresponding t o the curve f o r t he new s t ra in range a t t h e percentage of
l i f e used up, regardless of t h e s t r a i n l e v e l a t which t h e l i f e f r ac t ion
was consumed. Thus, the mater ia l may requi re i n i t i a l sof ten ing t o ap-
proach the necessary curve followed by hardening as it r i s e s along t h a t
curve, o r it may requi re i n i t i a l hardening followed by subsequent sof ten-
i ng f o r t h e same reason. mese r e s u l t s a r e remarkably i n t e r e s t i n g be-
cause they imply t h a t a t any condition of consumed l i f e based on the
l i n e a r damage ru l e , t he mater ia l seeks a s p e c i f i c s t r e s s l e v e l assoc ia ted
with t h e cyc l i c hardening or sof tening curve connected with i t s cur ren t
s t r a i n value and consumed l i f e f rac t ion . It i s i n t e r e s t i n g t o observe
from f igu res 42 and 43 t h a t , a t f r ac tu re , t he summation of t h e r a t i o s
i s qu i t e c lose t o uni ty , as indicated by the end points of t h e t es t
curves regard less of the sequence of loading.
Figure 44 shows the r e s u l t s of t e s t s of t h i s kind f o r a d i f f e r e n t
mater ia l - a t i tanium a l l o y - which i s b a s i c a l l y s t r a i n softening. Here
again, t he curves can be followed through f o r t he various t e s t s i nd i -
cated.
mater ia l previously discussed, bu t it w i l l be seen t h a t , when t h e change
The b a s i c behavior is qu i t e similar t o t h a t of a s t ra in-hardening
40
i s made i n s t r a i n leve l , t he new s t r e s s sought by the material does not
qu i t e reach the curve associated with t h e new s t r a i n leve l .
i n f igure 44(a) , i f t h e high s t r a i n i s appl ied f o r t he f i rs t 24 percent of
l i f e , t he curve follows segment AE.
produces a po in t E ' considerably below the curve CD a t t h e 25-percent-
l i f e level . Further cycl ing a t t h e lower s t r a i n l e v e l produces t h e
curve E'F, which never qu i t e reaches t h e curve CD. This behavior i s
cha rac t e r i s t i c of a l l t h e t e s t s shown i n t h i s curve and requi res f u r t h e r
invest igat ion.
For example,
Changing t o t h e lower s t r a i n l e v e l
A summary of a l l of t hese tests on t h e b a s i s of t he l i n e a r damage
r u l e i s indicated i n f igu re 45.
one of the t e s t s conducted.
high s t r e s s was appl ied f irst or last. Thus, f o r example, along the l i n e
f o r specimen L26, it i s seen t h a t t h e f i r s t 25 percent of l i f e w a s a t a
higher s t r a i n l e v e l because t h i s port ion of t he curve i s higher than the
remaining portion.
these l ines . It can be seen t h a t , f o r a l l t he t e s t s summarized, the cycle
r a t i o l i e s between approximately 0.8 and 1.3 and i s very c lose t o u n i t y
f o r most of t he t e s t s . Even t h e extremes a r e within t h e s c a t t e r band of
expected behavior i n t h i s type of test.
Each s e t of hor izonta l l i n e s represents
A s t e p i n t h e l i n e ind ica tes whether t he
Fracture i s indicated by the end point of each of
These t e s t s , therefore , would imply t h a t t h e l i n e a r damage r u l e i s
q u i t e accurate f o r these materials under the types of loading used. The
s ignif icance of these observations i n t h e l i g h t of w h a t has a l ready been
said regarding the development and propagation of a crack i s extremely
in te res t ing .
developing a very small crack r a t h e r than i n propagating the crack.
I n these tes ts t h e main por t ion of t he l i f e w a s used UP i n
It
~ 41
~
would appear, therefore , that a s long as the major por t ion of t he l i f e
i s devoted t o crack development the l i n e a r damage r u l e might be appl iable .
It may be noted t h a t t h i s f ea tu re of assuming the v a l i d i t y of a l i n e a r
damage r u l e within t h e crack i n i t i a t i o n s tage i s inherent i n t h e method
developed by Grover ( r e f . 24) f o r the ana lys i s of cumulative f a t igue
damage.
Aspects Relating t o Crack Propagation
I n the f a t igue behavior of a x i a l l y cycled small specimens, t he
propagation of a crack of engineering s igni f icance (ice., longer than
0.003 i n . ) does not c o n s t i t u t e a major port ion of t h e l i f e . I n other
cases, however, crack propagation might be much more s ign i f i can t . Let
us consider, therefore , what the crack growth curves should look l i k e
i f they are t o comply with the concept of t he l i n e a r damage ru le .
Figure 46 shows t h e idea l ized crack growth behavior i f t h e crack length
i s regarded as t h e only measure of mater ia l damage during cumulative
f a t i g u e t e s t s .
growth f o r two conditions of t e s t , assuming l i v e s of 100 and 1000
cycles. Curves OABC and OA'B 'C ' a r e t he idea l ized curves f o r crack
growth when the hor izonta l s ca l e i s p lo t t ed i n a c t u a l cycles.
curves a r e p lo t t ed on t h e b a s i s of percent of l i f e , as on the r i g h t of
t h e f igure , the assumption of a l i nea r damage ru l e , together with t h e
assumption t h a t crack length i s the only measure of f a t igue damage,
implies t h a t both curves become coincident.
75 percent of t h e hundred-cycle-life curve, OABC, while point A ' w i l l be
a t t h e 75 percent po in t of the thousand-cycle curve of
l a r l y points B and B' must coincide, and C and C ' must coincide.
On t h e l e f t of the f igure i s shown the idea l ized crack
If t h e
Thus, po in t A w i l l be a t
OAsBIC. Simi-
Since
42
t h e crack growth curves a r e of such a nature t h a t they coincide when
p lo t t ed on t he b a s i s of percent of l i f e , then it does not matter whether
one follows along the l i n e OA and then moves t o A ' a t a constant crack
length, continuing t h e t e s t from A' t o C', or whether one follows from
OA t o AC along t h e same curve. Each piece of curve t raversed between
any two of t he l e t t e r s indicated represents t he same f r a c t i o n of l i f e
c ons ume d . It can be seen, however, t h a t t h i s idea l ized behavior of t h e crack
A t least one reason i s indicated propagation behavior i s not possible.
i n f igu re 47.
occurs depends a t least t o a l imi ted ex ten t on t h e stress or s t r a i n t h a t
i s being applied. Obviously, f o r a low stress range or s t r a i n range, a
g rea t e r crack length w i l l be permitted before f r a c t u r e occurs. Thus, as
seen i n f igure 47, t he specimen with the 100-cycle l i f e , implying a high
s t r a i n range, w i l l f a i l a t , f o r instance, a r e l a t i v e crack length of 4
a t point B, while t he high l i f e specimen, implying a low appl ied strain
range, w i l l f a i l a t a l a r g e r crack length of 6 a t When p lo t t ed on
the b a s i s of percent l i f e t o f r ac tu re , it becomes impossible f o r t h e two
curves t o coincide a t a l l crack length l e v e l s because t h e r e i s a region
of crack length t h a t corresponds t o one of t he curves that does not even
e x i s t i n the other curve. A t be s t , therefore , t he kind of r e s u l t t h a t
can be obtained i s shown i n f igure 47(b). It is possible t h a t a small
region of t h e curve - perhaps from 0 t o A,A' - can correspond t o both
the curves OA and OA' of t h e c y c l i c - l i f e port ion of f igu re 47; beyond
t h i s point, however, some deviat ion between the two curves must Occur.
Once t h e two curves a r e not longer coincident, t h e l i n e a r damage r u l e
The permissible crack length i n a specimen before f r a c t u r e
C'.
43
cannot apply because, a t a given crack length, d i f f e r e n t percentages of
l i f e a r e used up depending on t h e s t r a i n l e v e l a t which the crack w a s
generated.
I n r e a l i t y the problem i s much more complicated than ind ica ted i n
f igu re 47. The main problem i s t h a t crack length cannot be regarded as
t h e s o l e measure of cumulative fa t igue damage. Among other important
f a c t o r s a r e r e s idua l s t r e s s (espec ia l ly i n the v i c i n i t y of a propagating
crack); meta l lurg ica l changes, pa r t i cu la r ly s t r u c t u r a l s t a b i l i t y ; t he
number of cracks present and t h e i r s t a t e of coalescence, e t c . But even
if we s t i l l continue t o analyze cumulative f a t igue damage on t h e s o l e
b a s i s of crack length, the l i n e a r damage r u l e i s inva l ida ted by f a i l u r e
of t h e crack-propagating curves t o coincide, as shown i n f i g u r e 48, when
p lo t t ed on a basis of percent l i f e . The bas i s f o r t h i s lack of coin-
cidence has a l ready been indicated e a r l i e r when discussing t h e crack
growth curves f o r t he polycarbonate r e s i n i n f igu re 38. It w i l l be
r eca l l ed t h a t these curves were indeed p lo t t ed on t h e b a s i s of percent
of l i f e and d id not coincide. The curve f o r t h e low s t r e s s l e v e l i nd i -
ca t e s t h a t t h e crack does not develop u n t i l much l a t e r i n l i f e , on a
percentage bas is . Uhce developed, the crack grows more r ap id ly when the
p l o t i s made on the basis of percent of l i f e . This f a c t i s i l l u s t r a t e d
i n f i g u r e 48(b). The curves OAB and OA'B'C' are r e a l i s t i c representa-
t i o n s of crack growth behavior f o r two s t r a i n leve ls . The crack growth
curve OAB f o r t he high s t r a i n i s s teeper a t the lower percentages of
l i f e , b u t terminates a t a lower f r ac tu re crack length. The low s t r a i n
curve i s f l a t t e r i n t h e early percentages of l i f e bu t becomes very s t eep
i n t h e te rmina l stages.
44
The s ignif icance of such behavior when viewed i n the l i g h t of t h e
l i n e a r damage r u l e can now be analyzed. Suppose, f o r example, t h e high
s t r e s s i s appl ied f i r s t along OA f o r a l i f e of 50 percent. A t t h i s
po in t a change is made t o t h e low s t r a i n range.
as the only c r i t e r i o n of f a t igue damage, assuming t h a t a crack of a given
length corresponds t o a given amount of damage regardless of what s t r a i n
l e v e l was used t o induce the crack, then changing t o t h e lower s t r a i n
l e v e l implies a move t o point A ' and completion of the t e s t t o f a i l u r e
along l i n e A ' B ' C ' . A t A ' , however, 75 percent of l i f e has a l ready
been used up; therefore , the l i n e
l i f e . The t o t a l l i f e f r a c t i o n used i s thus 0.5 plus 0.25 as ind ica ted
i n the tab le i n f igu re 48(b) , r e s u l t i n g i n a cycle r a t i o of 0.75.
If crack length i s used
A ' B ' C ' cons t i t u t e s only 2 5 percent of
Reversing the procedure and applying the low s t r a i n range f i r s t
t o point A ' produces a l i f e f r a c t i o n of 0.75. Changing t o t h e high
s t r a i n range implies a move back t o point
l i f e .
indicated i n the tab le . We see, therefore , t h a t t he order of appl ica-
t i o n of the s t r a i n s may se r ious ly a f f e c t t h e summation of t h e cycle
r a t i o s .
range produces a cycle r a t i o l e s s than uni ty; applying the low s t r a i n
range f i r s t followed by the high s t r a i n range produces a cycle r a t i o
g rea t e r than unity.
A, leaving 50 percent of
Thus, a cumulative l i f e f r a c t i o n of 1 . 2 5 may be obtained as again
Applying the high s t r a i n range f i rs t followed by the low s t r a i n
Experience has v e r i f i e d t h i s conclusion. Many t e s t s have been con-
ducted i n which s t e p functions i n s t r e s s or s t r a i n have been applied.
Two-step t e s t s have been the most common. These t e s t s , i n general ,
have ve r i f i ed t h e conclusion t h a t t he order of appl ica t ion i s important
45
and t h a t when a low s t r a i n range i s applied f i r s t l e s s damage i s in -
curred than when t h e high s t r a i n range i s f i r s t applied. We have con-
ducted many such t e s t s i n o w own laboratory and t h i s subject has been
touched upon i n some of our e a r l i e r publ icat ions ( r e f s . 25 and 26) .
An important deviat ion from t h i s conclusion has, however, been
observed i n the t e s t i n g of notched specimens.
found t h a t applying the high s t r e s s range f i r s t t o such a specimen
r e s u l t s i n a higher cycle r a t i o summation than applying the low s t r e s s
range f i r s t . There a r e severa l reasons f o r t h i s observation. Perhaps
the most important i s the f a c t t h a t when a notch or a crack i s present ,
important r e s idua l s t r e s s e s are retained when t h e cyc l i c load range i s
reduced from an i n i t i a l high value t o a low value.
cyc l i c t e s t s run from zero t o tension only, wnen applying the high s t r e s s
range f i r s t , a r e s i d u a l compressive s t r e s s i s produced a t the crack t i p
on unloading. A considerable number of cycles a t a lower s t r e s s or
s t r a i n range would then be required t o overcome the r e s idua l s t r e s s
f i e l d produced by t h e p r i o r h i s tory .
Weibull ( r e f . 27) and others , and the delay time i n continued cracking
has been s u b s t a n t i a l when the high s t r e s s i s f i r s t applied. This observa-
It has, i n f a c t , been
For example, f o r
This behavior has been s tudied by
1 - t i o n makes it c l e a r why crack length i t s e l f cannot be considered as the
only c r i t e r i o n f o r t h e exis tence of damage. A t l e a s t t he question of
r e s i d u a l s t r e s s f i e l d a t t he crack t i p must be included.
There is, i n addi t ion , another reason f o r t he ordering e f f e c t of
load appl ica t ion observed i n notched specimens. This reason w i l l be
i l l u s t r a t e d with the use of f igu re 49. It w i l l a l s o be shown t h a t , i n
two-level cumulative damage t e s t s , one may reach a d i f f e r e n t conclusion
46
as t o the r e l a t i v e damage done by applying the high or low s t r e s s f i r s t ,
depending on whether a notched or unnotched specimen i s being considered.
This difference i n behavior becomes obvious after a comparison i s made
of f igures 48(b) and 49(b).
Consider, f o r example, t he crack growth behavior of a specimen with
This crack may have been generated
Figure 49(a) shows how t h i s un i t
an i n i t i a l crack length of one uni t .
by machining o r by p r i o r load cycling.
crack length could be produced i n an i n i t i a l l y unnotched specimen under
two a r b i t r a r y conditions of s t r a i n cycling.
same as those shown i n f igu re 48(a) f o r t he crack growth of unnotched
specimens.
mens w i t h t he un i t length crack a r e 50 and 250 cycles, respect ively.
These may now be regarded as notched specimens having known fa t igue
l i v e s (when cycled a t the same s t r a i n used t o produce the notch) of
50 and 250 cycles.
damage i s the remaining crack growth behavior of these specimens repre-
sented as a funct ion of percent of l i f e t o f a i l u r e , or percent of l i f e
remaining a f t e r the crack has been generated.
i n f igu re 49( ' b ) . The crack growth curves a r e shown as AB and A ' B ' C ' ,
i n which AB now l i e s e n t i r e l y below A ' B ' C ' . If an ana lys i s were made
of t he e f f e c t of order of appl ica t ion of s t r a i n i n a manner analogous t o
t h a t discussed i n connection with f igu re 48, it would be seen t h a t t h e
summation of the cycle r a t i o s would be g rea t e r than un i ty i f the high
s t r a i n were applied f i r s t i n notched specimens.
l a t e d t o any aspect of r e s idua l s t r e s s bu t i s simply t h e consequence
of the crack growth behavior of a notched specimen ( i . e . , using the
These growth curves are the
Let us assume t h a t the remaining l i v e s of these two speci-
What i s important t o t h i s discussion of cumulative
This behavior i s shown
This r e s u l t i s not r e -
47
remaining l i f e a f t e r the crack has reached an appreciable s i z e i n an
i n i t i a l l y unnotched specimen as the basis f o r ca lcu la t ing cycle r a t i o s ) .
Obviously t h i s e f f e c t may a l s o be due t o r e s idua l stresses and metal lur-
i c a l e f f e c t s , but it i s of i n t e r e s t i n understanding apparent inconsis-
t enc ie s i n observations regarding the e f f e c t of order i n t h e appl ica t lon of
stresses or Strains .
The subject or cumulative damage i s covered by a vas t l i t e r a t u r e .
Further information on t h i s subjec t may be found i n references 21 t o 36.
APPLICATIONS INVOLVING STRAIN CYCLING DATA
Relation Between Axial and Bending Fatigue
The s igni f icance of t he s tud ie s of mater ia ls i n s t r a i n cycl ing
w i l l now be i l l u s t r a t e d by a very simple appl ica t ion t h a t can a l s o be
extended t o more complicated problems. I n t h i s appl ica t ion w e have
concerned ourselves with t h e r e l a t ion between the a x i a l f a t igue be-
havior of mater ia ls and t h e i r bending f a t igue behavior.
It i s well known t h a t metals have longer l i v e s when t e s t e d i n
bending f a t igue f o r t h e same nominal stress l e v e l than they have i n
a x i a l f a t igue . This f a c t i s i l l u s t r a t e d i n f igure 50. Here the
f a t i g u e l i v e s of specimens of 4130 s t e e l a r e shown obtained by two
d i f f e r e n t methods. The lower curve is t h e sa tu ra t ion s t r e s s range
p l o t t e d aga ins t t he cyc l ic l i f e when the s t r a i n range i s a x i a l l y applied
and, therefore , approximately uniform over the e n t i r e cross sect ion.
The upper curve r e f e r s t o r o t a t i n g beam specimens.
i s t h e nominal value obtained from the e l a s t i c bending moment formula
S = Mc/I. Obviously the re i s a considerable difference between these
two curves.
The s t r e s s p lo t t ed
48
A question a r i s e s as t o the reason f o r t h i s difference. One reason
obviously i s t h a t t h e formula, which assumes e l a s t i c i t y , i s not accurate ,
espec ia l ly i n the low-cycl ic- l i fe range where the s t r a i n s a r e wel l i n t o
the p l a s t i c range.
t h e a x i a l and bending f a t igue r e s u l t s would be the same i f the a c t u a l s t r e s s
or st rain i n the r o t a t i n g beam specimen were known.
Thus, i n t h i s example we s h a l l inves t iga te whether
A s s h a l l be indicated l a t e r , t he question of r o t a t i n g beam bending
i s much more complicated than appears upon f i rs t examination, and we
s h a l l therefore look f i rs t a t a similar problem which is in t roductory t o
the study of r o t a t i n g beams. Instead of considering a specimen which i s
f lexed during ro t a t ion , l e t us consider f i r s t a round specimen subjected
t o simple reverse f lexure. Imagine t h e specimen t o be bent back and f o r t h
about a f ixed ax is . The b a s i s for t he ana lys i s i s indicated i n f igu re 51.
If we t r y t o solve the problem by s e l e c t i n g a bending moment and de te r -
mining the corresponding surface s t r a i n range, we encounter d i f f i c u l t i e s
i n solving t h e complicated nonlinear equation t h a t r e s u l t s . To avoid
these d i f f i c u l t i e s we s e l e c t ins tead the surface s t r a i n (which, i n tu rn ,
es tab l i shes the l i f e ) and seek t o determine the bending moment t h a t w i l l
produce the se lec ted surface s t r a i n . This can r ead i ly be done as i l l u s -
t r a t e d i n f i gu re 51. For example, if w e s e l e c t a surface s t r a i n of 0.0105
and make the assumption, commonly made i n bending ana lys i s , tha t plane
sect ions remain plane, t h e s t r a i n i s known a t every poin t i n the specimen
cross section. A t any v e r t i c a l d i s tance x from the center of t h e speci-
men, the s t r a i n i s x/R (0.0105), as shown i n f igu re 51(a). Knowing the
s t r a i n range a t each locat ion, and having ava i l ab le a cyc l i c s t r e s s - s t r a i n
curve t h a t r e l a t e s the s t r a i n range t o t h e s t r e s s range, we may then
49
determine the s t r e s s range a t each location i n t h e cross sec t ion . This
i s i l l u s t r a t e d i n f igu re 51(b). For t h e materials se l ec t ed t h e sur face
s t r e s s i s 94 800 p s i and t h e s t r e s s e s a t a l l o ther d i s tances from t h e
a x i s toward t h e center a r e displayed by t h e curve shown. The cont r ibu t ion
t o t h e bending moment associated with each stress can be determined from
the product of t h e area over which the s t r e s s a c t s and t h e d is tance from
the bending ax is . On car ry ing out
t h e ind ica ted in tegra t ion , we conclude f o r t h e example shown that t h e
bending moment i s 113 000 inch-pounds.
we can then determine from the equation
ing s t r e s s would have been fo r t h i s bending moment.
poss ib le t o make a p l o t of t he applied nominal s t r e s s aga ins t p red ic ted
l i f e and compare it with the experimental data.
This i s i l l u s t r a t e d i n f i g u r e 51(c) .
Once t h e bending moment i s known,
what t h e nominal bend-
I n t h i s way it i s
S = Mc/I
Results f o r s eve ra l mater ia l s a r e shown i n f i g u r e 52. The continu-
ous l i n e s r e f e r t o pred ic t ions determined by t h e described procedure,
and t h e experimental data points a r e f o r t e s t s conducted t o determine
t h e v a l i d i t y of t h e predictions.
t he two s t e e l s and two aluminum a l loys i s q u i t e good.
pred ic t ions a r e somewhat conservative. The reason f o r t h i s i s t h a t t h e
pred ic t ions a r e based on a x i a l fa t igue data i n which t h e crack growth
s t age is , as a l ready discussed, a r e l a t i v e l y small part of t h e t o t a l
l i f e .
t he assumption t h a t failure occurs soon a f t e r t h e crack starts i s some-
what conservative f o r t he bending t e s t s , and t h e pred ic t ions a r e a l s o
conservative. Another reason for thediscrepancy, perhaps a minor f ac to r ,
i s t h e assumption t h a t , throughout the l i f e , each point i s a t t h e satura-
It can be seen t h a t t h e agreement f o r
I n general, t he
I n t h e bending t e s t s t h e crack growth s tage i s longer. Therefore,
50
t i o n s t r e s s l e v e l assoc ia ted with i t s s t r a i n value. This i s a reasonable
assumption f o r t he outermost f i b e r where t h e s t r a i n i s a maximum ( r e c a l l -
i ng that sa tu ra t ion i s achieved e a r l y i n t h e cyc l i c l i f e when t h e e n t i r e
s ec t ion i s subjected t o the same s t r e s s and recognizing a l s o that t h e l i f e
i s t o be governed by t h e s t r a i n a t t h e outermost f i b e r ) . The assumption
i s not va l id , however, a t po in ts of l e s s e r d i s tances from the n e u t r a l a x i s ,
which a re subjec t t o lower s t r a i n s , and f o r which a l a r g e r number of cycles
a r e required t o e s t a b l i s h sa tu ra t ion s t r e s ses .
t h e discrepancy i s t h e unknown e f f e c t of s t r a i n grad ien t on l i fe .
a x i a l s t r a i n cycling t e s t s , the s t r a i n i s uniform over t h e e n t i r e sec t ion ,
whereas i n t h e bending t e s t s , only a very small region a t t h e outermost
rad ius i s a t t h e maximum s t r a i n , while t h e remainder of t h e sec t ion i s a t
lower s t r a i n . Nevertheless, desp i t e t he many f a c t o r s t h a t could cause
S t i l l another reason for
In .the
discrepancy, t h e agreement between predic t ions and experiment shown i n
f i g u r e 52 i s seen t o be very good.
We can now proceed t o examine t h e more complicated case of r o t a t i n g
bending.
f l exura lbend ing lends i t s e l f so e a s i l y t o ana lys i s i s t h a t once t h e su r -
face s t r a i n has been assumed, a l l s t r e s s e s and s t r a i n s i n t h e cross
sec t ion are immediately es tab l i shed , s ince they a r e a l l i n phase, reach-
ing a maximumamplitude a t t h e same t i m e .
t i o n t h a t develops when considering r o t a t i n g bending i n con t r a s t t o
f l e x u r a l bending. Consider t h e two elements A and B shown i n
f i g u r e 53(b).
f o r t h as i n f l e x u r a l bending, these two elements, being a t t h e same d i s -
tance from t he bending a x i s , a .- a would have exac t ly t h e same s t r e s s .
Before doing so , l e t us recognize t h a t t h e reason t h e case of
Figure 53 shows t h e complica-
I f t h e cross sec t ion were simply being bent back and
51
In r o t a t i n g bending, however, the problem develops because of a phase
r e l a t i o n , with the r e s u l t t h a t A and B a r e not a t the same s t r e s s
l eve l , unless t h e e n t i r e cross sect ion i s e l a s t i c . If we were t o analyze
t h i s problem i n the same manner as we d id f o r f l e x u r a l bending, assuming
a s t r a i n a t the surface of E ~ , t he s t r e s s a t points A and B would
be equal and would be obtained from a knowledge of t h e s t r a i n , as shown
i n f i g u r e 53(c). If t h e cross sect ion were not ro t a t ing , A and B
would always be a t t h e same dis tance from the n e u t r a l axis, as shown i n
f igu re 53(b) , and no question would a r i s e regarding t h e value of s t r e s s
i n these elements.
c l e a r t h a t points A and B w i l l u l t imately reach a s t r a i n l e v e l Em
when they achieve the maximum distance from the strain axis , bu t t h a t a t
t h e i n s t a n t considered i n the f igure , point B has a l ready passed i t s
maximum s t r a i n , while po in t A i s approaching i t s maximum. The s t r e s s e s
a t each of these points must be determined from the hys te res i s loop
( f i g . 5 3 ( d ) ) .
s t r a i n em t o be reached by both these points i s shown as um, obtained
from f igu re 53(c) f o r t h e maximum s t r a i n
points A and B. To obtain the current s t r e s s e s , however, we must
determine the s t r e s s when the s t r a i n i s E. Therefore, r e f e r r i n g again
t o f i g u r e 53(d) , we can move out a distance
loca t ing the points i den t i f i ed as UA and uB. These a r e t h e s t r e s s e s
a t t h e two elements A and B. Thus, even though A and B a r e a t
t he same d is tance from the s t r a i n axis, t h e i r s t r e s s e s a re , i n f a c t ,
q u i t e d i f f e r e n t a t t h e i n s t a n t considered, and i n obtaining t h e bending
moment allowance must be made f o r the a c t u a l s t r e s s present i n each
element.
When w e consider r o t a t i n g bending, however, it i s
I n t h i s f i gu re the s t r e s s corresponding t o the maximum
ern experienced by both
E from the v e r t i c a l ax i s
52
Further r e f l e c t i o n regarding t h i s problem indica tes t h a t t he b e s t
approach f o r ana lys i s i s t h a t indicated i n f igu re 54. The loading axis xx
is the axis about which the bending moment is applied. If one considers
t he strains a t the individual loca t ions i n the element, it can be seen t h a t
t he cross sec t ion bends about a d i f f e r e n t axis cp
with t h e xx axis .
t r i bu ted by a l l the elements about an a r b i t r a r y ax i s r e l a t i v e t o the d i s -
placement ax is , and expressing the f a c t t h a t t h e bending moment about t h e
axis normal t o t he loading axis
ing moment has no component about t h i s normal axis) , we can determine t h e
angle
the figure.
provided i n a forthcoming NASA publication.
two axes has been determined, the complete s t r e s s and s t r a i n d i s t r i b u t i o n
can be computed, from which t h e bending moment f o r an assumed value of
maximum surface s t r a i n can be establ ished. From t h e assumed value of
surface s t ra in , l i f e can be determined from f igu re 53( a ) .
aa, which makes an angle
By wr i t i ng an expression f o r t he bending moment con-
xx must be zero ( s ince t h e appl ied bend-
cp between t h e def lec t ion ax i s and t h e loading axis, as shown i n
The computations a re somewhat complex, and d e t a i l s w i l l be
Once the angle between t h e
A comparison of t he predicted r e s u l t s and experimental data i s shown
i n f igu re 55.
the experimental data a r e indicated by t h e symbols.
qu i te good, c e r t a i n l y much b e t t e r than t h e agreement shown i n f igu re 50.
Thus, it can be concluded t h a t one of t h e main reasons why bending and
axial fa t igue do not agree i s t h a t we do not , i n r e a l i t y , know what the
t r u e s t r a i n i s on t h e surface f o r a given appl ied bending moment i n ro -
t a t i n g bending. When t h e proper surface s t r a i n i s determined by an ana-
l y s i s of the type ind ica ted here, g r e a t l y improved r e s u l t s a r e obtained.
The curves a r e the predict ions based on a x i a l f a t igue data;
The comparison i s
53
The problem discussed here is obviously an extremely elementary one,
bu t t h e bas i c approach involving t h e use of t he cyc l i c s t r e s s - s t r a i n
curve, can be extended t o much more d i f f i c u l t and more general problems.
Rule of Thumb Relating S t r a in Range t o Cyclic L i f e
I n an e a r l i e r publ icat ion (ref. 14) it' was observed t h a t , as a f i rs t
approximation, most mater ia ls w i l l survive approximately 10 000 cycles
of appl ica t ion of a s t r a i n range of 1 percent.
examined t h i s r u l e of thumb i n comparison with others and has concluded
tha t an even b e t t e r r u l e i s one suggested by Peterson (ref. 38) that
most mater ia ls can withstand 1000 cycles of a s t r a i n range of 2 percent
p r i o r t o f a i l u r e . Because of the large amount of data r ecen t ly developed
i n our laboratory r e l a t i n g l i f e t o s t r a i n range, it i s appropriate t o
reexamine both these rules .
Morrow ( r e f . 37) has
The r e s u l t s f o r t h e 29 mater ia ls previously used i n our discussion
are shown i n f igu re 56. Here a l l the data a r e shown i n a p l o t of s t r a i n
range aga ins t l i f e , regardless of the values of t e n s i l e s t r eng th and
duc t i l i ty of t h e ind iv idua l materials. It i s seen that i n t h e low-l i fe
range the re i s considerable spread i n t h e points because of t h e l a rge
range i n d u c t i l i t i e s of t he materials; i n t h e very long- l i fe range a
spread again develops because of t he l a rge range i n t h e s t r eng th l eve l s
of t h e mater ia ls tes ted . I n t h e range between 1000 and 10 000 cycles,
the spread narrows considerably, and each of t he two ru l e s c i t e d above
presents a reasonable approximation of a c t u a l mater ia l behavior. (The
very discrepant data points i n the low-cyclic-l ife range r e f e r t o the
powder-metallurgy produced beryllium, which probably d i d not reach i t s
f u l l p o t e n t i a l i n fa t igue . ) If a r i g i d choice had t o be made between
54
t h e merits of t h e two ru l e s , it would probably be i n favor of t h e 2 percent
r u l e a t 1000 cycles, bu t both ru l e s a r e usefu l f o r e s t ab l i sh ing t h e approxi-
mate behavior i n the range of l i f e between 1000 and 10 000 cycles.
FUNDAMENTAL ASPECTS OF FATIGUE
A s has a l ready been indicated, it i s not t h e in t en t ion here t o sum-
marize the s t a t e of knowledge of the f a t igue mechanism, s ince t h i s subjec t
i s not only too broad, bu t a l s o because many good summaries have r ecen t ly
appeared.
cussed from an engineering point of view a l s o have some fundamental s i g -
nif icance t h a t bear summarization. I n the following discussion we s h a l l
l i m i t ourselves pr imari ly t o r e s u l t s that have a l ready been mentioned i n
some other connection, introducing only a few add i t iona l f ea tu re s needed
t o round out t h e in t e rp re t a t ion of t hese r e su l t s .
However, some of t h e aspects of the subjec t that have been d i s -
Cyclic S t r a in Hardening and Softening
The f i rs t f ea tu re i n the f a t igue process t h a t we have observed i n
our discussion i s t h e change i n s t r e s s range required t o maintain a given
cyc l i c s t r a i n range.
such cycl ic hardening or sof ten ing a r e summarized i n t a b l e 111.
we have noted, as i n f igu re 8, t h a t such changes i n s t r e s s a r e manifesta-
t i o n s of microstructural changes. Since these changes a r e an ind ica t ion
of the ra ther la rge possible changes i n t h e y i e ld point , it can be seen
that t h e i n i t i a l y ie ld point i s of r e l a t i v e l y l i t t l e s ign i f icance i n
governing t h e fa t igue behavior when l a rge p l a s t i c s t r a i n s a r e involved.
The v a l i d i t y of t h i s implicat ion has been demonstrated by showing t h a t
the fat igue l i f e i n the low and intermediate cyc l i c l i f e range depends
very l i t t l e on the i n i t i a l y i e ld point.
The f ea tu res of major s ign i f icance assoc ia ted w i t h
F i r s t ,
We have a l s o ind ica ted t h a t
.
( 2 ) Tends t o wipe out i n i t i a l y i e l d poin t
55
sa tu ra t ion hardening or softening occurs r e l a t i v e l y e a r l y i n t h e l i f e of
a specimen, as i l l u s t r a t e d i n f igures 15 (c ) and 1 6 ( c ) , as long as t h e
s t r a i n range is maintained constant during t h e complete h i s t o r y of load-
ing, When changes a r e introduced i n t h e s t r a i n range during t h e course
of t he l i f e of a specimen, these changes may be accompanied by l a rge
changes i n hardening or softening, the material seeking a new stress
l eve l , which would place it on the hardening or sof ten ing curve asso-
c i a t ed with t h e new s t r a i n leve l . The po in t on t h e curve it seeks de-
pends roughly on t h e sum of t h e cycle r a t i o s of t h e p r i o r h i s tory . Thus,
as ind ica ted i n item (4) of t a b l e 111, we conclude t h a t changes i n stress
a r e governed approximately by a l inea r l i f e - f r a c t i o n ru le . Limited ev i -
dence of t h i s behavior i s shown i n f igures 42 - 44, b u t more data a r e
needed f o r more general subs tan t ia t ions . F ina l ly , we observe, as i n
item (5) of t a b l e 111, t h a t cyc l i c hardening and sof ten ing may appre-
c i a b l y a l t e r t h e macroscopic s t r e s s d i s t r i b u t i o n i n an engineering
s t r u c t u r e , s ince t h e s t r e s s d i s t r i b u t i o n depends on the cyc l i c s t r e s s -
s t r a i n curve, which i n t u rn i s influenced by t h e hardening or softening.
56
An example of t h i s e f f e c t has been discussed i n connection with the bend-
ing problem, where it has been indicated t h a t computations using t h e
cyc l i c s t r e s s - s t r a i n curves lead t o improved determination of t he a c t u a l
sur face stresses and s t r a i n s , which makes possible a more accurate deter-
mination of l i f e .
Microstructural Events i n Fatigue
The microstructural events assoc ia ted with hardening and sof tening,
and which lead t o the eventual development of a crack t h a t can then s t a r t
propagating i n subsequent cycling, are out l ined i n table I V . A s i nd i -
TABLE I V . - MICROSTRUCTURAL EVENTS I N FATIGUE
(1) S l i p on a f e w well-defined planes tends t o produce
in t e rna l d i scon t inu i t i e s .
( 2 ) Substructure formation helps maintain cont inui ty; it
r e s u l t s , however, i n more centers of d i sorder .
(3) S l i p may a f f e c t t h e nature of t he microstructure,
f o r example, break up dispersed p a r t i c l e s ( s t r a m sof ten ing) .
( 4 ) S l i p not completely reversible, r e s u l t s i n i n t e r n a l
and surface d i scon t inu i t i e s .
cated i n item (l), t he most important f a c t o r t o bear i n mind i n connection
with fat igue i s t h a t , i n c r y s t a l l i n e materials, p l a s t i c deformation of
any one c r y s t a l can f o r t h e most pa r t take place only i n a f e w , well-
defined d i rec t ions assoc ia ted with t h e s l i p planes. Thus, while t h e
equations of compat ib i l i ty ( those equations t h a t r e s u l t from t h e as-
sumption t h a t the material remains continuous) s t i p u l a t e a s t r a i n d i s -
t r i b u t i o n i n - a l l d i r ec t ions , such s t r a i n s may not be poss ib le when t h e
.
57
ava i lab le d i r ec t ions of s l i p a r e l imi ted . Thus, if t h e s t r a i n develops
only as a r e s u l t of s l i p on such a l imi ted number of s l i p planes, there
develops a tendency t o c rea t e poin ts of mater ia l d i scont inui ty . Naturally,
t he more grainsthere a r e per un i t area, each grain being d i f f e r e n t l y
or iented so t h a t i t s s l i p planes a r e i n d i f f e r e n t d i rec t ions , t he eas i e r
it becomes t o s a t i s f y the compatibi l i ty requirements without producing
local ized poin ts of d i scont inui ty , For t h i s reason, as ind ica ted i n
item ( 2 ) o f t a b l e IV, and demonstrated i n f igu re 8, a substructure
develops within t h e individual grains. That is, the gra ins subdivide
i n t o smaller regions, each having a s l i g h t l y d i f f e r e n t o r i en ta t ion r e l a -
t ive t o the adjacent regions, s o t h a t i t s s l i p planes a r e i n a d i f f e ren t
d i rec t ion . Thus there is es tab l i shed a grea te r ease of obtaining a r e -
quired p l a s t i c s t r a i n d i s t r ibu t ion . Such realignment of t h e subgrains
i s not, however, accomplished without paying a pr ice . This p r i ce i s i n
the form of grea te r disorder i n the subgrain boundaries ( r e f . 39), which
eventually may serve as nuc le i f o r f a t igue cracking, b u t a t least i n the
e a r l y stages must be regarded as bene f i c i a l .
While cyc l i c s t r a i n hardening can readtilybe in t e rp re t ed according
t o accepted concepts involving the exhaustion of planes of easy s l i p ,
t he pile-up of d i s loca t ions , or any other mechanism t h a t requi res more
force t o cause a given amount of s l i p a f t e r s l i p has a l ready occurred
previously, s t r a i n sof tening i s somewhat harder t o understand physically.
Item (3) o f t a b l e I V ind ica tes one mechanism f o r s t r a i n sof ten ing t h a t
i s appl icable t o dispersion-strengthened mater ia l s (mater ia l s t h a t gain
t h e i r s t rength as a r e s u l t of a d ispers ion of very f i n e hard p a r t i c l e s ,
such as carbides i n s t e e l and oxides i n aluminum). McEvily and Boettner
( r e f . 40) have found, f o r example, t h a t i n an aluminum a l l o y which gains
i t s s t r eng th i n t h i s manner, cyc l i c s t r a in ing caused the hard, b r i t t l e
p a r t i c l e s t o be broken up whenever s l i p occurred along a plane i n t e r s e c t i n g
them. Eventually, the p a r t i c l e s become so small t h a t they become unstable
i n t h e i r matrix. When the p a r t i c l e s become small enough, t h e r a t i o of
t h e i r sur face t o volume becomes very high, and they become unable t o r e -
s is t d i s so lu t ion i n t o the surrounding matrix. A s t h e p a r t i c l e s dissolve,
t h e region becomes s o f t because of the removal of t h e hardening agents,
thus explaining a mechanism of cycl ic s t r a i n softening. Other explana-
t i o n s a re , of course, required f o r other systems. The d i s t r i b u t i o n of
f o r example, be unstable when t h e material i s subsequently subjected
t o cyc l i c p l a s t i c s t r a i n . A new d i s loca t ion s t r u c t u r e and d i s t r i b u t i o n
may thus develop t h a t more r e a d i l y lends i t s e l f t o s l i p , producing s t r a i n
sof tening.
The f i n a l item, item (4) of tab le I V , r e f e r s t o a mechanism whereby
a crack i s eventual ly developed as a r e s u l t of cyc l i c r eve r sa l of s t r a i n : It
has been proposed by Wood ( r e f . 41), among o thers , and i s i l l u s t r a t e d i n f i g -
u re 57.
of s l i p of any one s l i p plane.
The concept i s associated with t h e (a t l e a s t p a r t i a l ) i r r e v e r s i b i l i t y
I n f igu re 57(a), s l i p takes place along one
s l i p plane when t h e load i s t e n s i l e as shown. Reversing t h e load does
not reverse the s l i p on the iden t i ca l plane; a t l e a s t a part of t h e re -
versed s l i p takes place on an adjacent p a r a l l e l plane, producing a s m a l l
notch a t t h e sur face (or i n the i n t e r i o r a t a gra in boundary i n some
cases) , as shown i n f igu re 57(b).
t i n u i t y , as shown i n f igures 57(c) and ( d ) , eventual ly developing a crack
Further cycling enlarges the discon-
59
of macroscopic proportions t h a t responds t o engineering formulas f o r
s t r a i n concentration i n t h e v i c i n i t y of a notch, thus f u r t h e r aggravating
s t r a i n loca l i za t ion . A notch such as t h a t shown i n f igure 57(d) has been
ca l led an in t rus ion , because of t he absence of material; when the s l i p i s
of such nature as t o form an excess of ma te r i a l j u t t i n g out from t h e s u r -
face, it i s ca l l ed an extrusion.
i l l u s t r a t e d i n f igu re 13.
Examples of such extrusions have been
Development of Microcracks
Further aspects of t h e development of microcracks a r e out l ined i n
t a b l e V. As noted i n item (l), such microcracks may develop a t regions
TABLE V. - DEVELOPMENT OF MICROCRACKS
(1) Influenced by i n i t i a l surface topography, in t rus ions ,
extrusions, inclusions, f l a w s , g ra in boundaries,
subgrain boundaries, twin boundaries, e t c .
( 2 ) I n i t i a l d i r ec t ion p a r a l l e l t o s l i p planes; s tage I
(3) Slow growth; l imi ted within ind iv idua l grains; may
stop a t grain boundary o r other impediment (non-
propagating crack)
(4) Main f r a c t i o n of l i f e used up i n t h i s s tage, except
a t extremely high s t r a i n and low l i f e
(5) Mater ia ls of high strain capac i ty . (high d u c t i l i t y
f o r p l a s t i c s t r a i n , high t e n s i l e s t rength f o r
e l a s t i c s t r a i n ) r e s i s t e a r l y development of
microcracks
( 6 ) Tends t o follow cumulative l i n e a r l i f e f r a c t i o n r u l e
60
of strain concentration associated with in t rus ions and extrusions a l ready
discussed, bu t they may a l s o develop a t other regions of d i scont inui ty
such as surface i r r e g u l a r i t i e s , inclusions o r o ther i n t e r n a l flaws, a t
grain boundaries, subgrain boundaries, twin boundaries, or wherever a
d iscont inui ty of s l i p o r s l i p d i r ec t ion occurs. As a l ready indicated
i n f igure 57, the i n i t i a l d i r ec t ion of the crack i s l i k e l y t o be parallel
t o a s l i p plane, which i s fu r the r i l l u s t r a t e d i n f igu re 58, taken from
a recent repor t of Schijve ( r e f . 42) .
by the f ine l i nes , and the crack by the coarse l i ne . Note t h a t i n any
one grain the crack i s p a r a l l e l t o t he s l i p plane, and that as the crack
moves t o an adjacent gra in it changes i t s d i r ec t ion s o as t o be p a r a l l e l
t o t h e s l i p planes of t he new grain.
however, only f o r t he e a r l i e s t s tages of cracking.
l a t e r stages i s t r a n s c r y s t a l l i n e , with r e l a t i v e l y l i t t l e regard for the
d i rec t ion of t h e s l i p planes.)
t h e crack i s confined l a r g e l y t o an ind iv idua l grain, o r t o a few
adjacent grains , has been c l a s s i f i e d as s tage I by Forsyth and by
others . As indicated i n item ( 3 ) , of t a b l e V, t he growth of such a
s ingle-grain confined crack i s qu i t e slow. When it reaches a gra in
boundary it may, i n f a c t , be stopped if t h e o r i en ta t ion of the s l i p
planes of the adjacent grain i s such as t o f a i l favoring propagation,
Many examples have been c i t e d i n t h e l i t e r a t u r e of so-cal led nonpropa-
ga t ing cracks, a t l e a s t some of which may be explained on the basis of
t he favorable o r i en ta t ion of adjacent gra ins s o as t o prevent crack
extension once t h e gra in boundary or other impediment i s reached.
The s l i p planes are del ineated
(Such pa ra l l e lnes s i s typ ica l ,
Cracking i n the
The e a r l y s tages of cracking, wherein
Item ( 4 ) of t a b l e V r e f e r s to t h e conclusion drawn from f igu re 39
61
t h a t t he major port ion of t h e l i f e i s used up i n i n i t i a t i n g t h e crack.
Because t h i s conclusion may appear t o b e i n contradict ion t o many cur ren t
concepts t h a t f a t igue i s predominantly the process of propagating a
crack, it i s important t o c l a r i f y the s ign i f icance of t h i s item i n order
t o avoid undue dispute i n cases where no r e a l c o n f l i c t e x i s t s .
w e must recognize t h a t t h e conclusion r e l a t e s t o cyc l i c l i v e s g rea t e r
than about 500 cycles. In the very low-life-range, figure 39 shows t h a t
crack propagation consumes, i n any case, t he major p a r t of t he l i fe . A t
a l i f e of 100 cycles, f o r example, only about 40 percent of t h e l i f e i s
required t o i n i t i a t e t he crack under consideration, and about 60 percent
i s required t o propagate it t o f i n a l f r ac tu re . Secondly, we must recog-
n ize t h a t we a r e dea l ing with o r ig ina l ly smooth specimens, wherein
i n i t i a l l y the re a r e no gross s t ra in-concentrat ing conditions. If a
sharp d i scon t inu i ty is i n i t i a l l y present, t h e crack may develop and
start t o grow much e a r l i e r i n t h e l i f e . That t h i s is t h e case has been
i l l u s t r a t e d i n f igu re 14, where t h e specimen i n i t i a l l y contains a very
sharp notch, and it can be seen t h a t a crack of about 0.0025 inch i s
de tec ted within about 1 percent of the l i f e almost over t h e e n t i r e l i f e
range f o r t he specimen. Furthermore, we must recognize t h a t we are con-
cerned wi th small specimens of 1/4-inch c i r c u l a r cross sec t ion , avoiding
s t ress -cor ros ion conditions or other h o s t i l e environments t h a t might be
t y p i c a l of se rv ice appl ica t ions . Final ly , and perhaps of t h e g r e a t e s t
importance, we must recognize that when w e talk about a crack w e mean
one a t l e a s t equal i n s i z e t o t h e na tura l imperfections present on the
sur face of engineering materials. Any smaller cracks w i l l have an in -
s i g n i f i c a n t e f f e c t compared with the imperfections a l ready present.
F i r s t ,
62
Undoubtedly, "cracks" developed by f a t igue can be detected i n highly
polished laboratory samples e a r l i e r than shown i n f igu re 39 if use i s
made of the very high magnifications associated with e lec t ron micros-
copy or of a high-power op t i ca l microscope together with t ape r sect ion-
ing. Many inves t iga tors have observed "cracks" by such techniques very
e a r l y i n the l i f e of a specimen, and t h e i r observations a r e not i n d i s -
pute.
of f igure 39 were applied were associated with engineering problems
r a the r than with t h e i r i n t e rp re t a t ion from the standpoint of physics.
When considering notch s e n s i t i v i t y , which w a s t he f i rs t appl ica t ion
involved, it would seem reasonable t h a t we would want t o consider cases
only when the crack i s la rge enough t o a f f e c t t h i s s e n s i t i v i t y i n an
engineering sense, as already discussed. I n t h e second appl icat ion,
t h a t of considering t h e crack growth l a w , we were concerned with w r i t -
ing an equation f o r crack growth during i t s "measurable" s tage - whether
by " s t r i a t ions" , "r ipples" , o r otherwise. We needed some boundary con-
d i t i o n s for i n t eg ra t ing the equations, and wanted t o ensure t h a t t h e
corresponding values of 2, and No were within t h e range of v a l i d i t y
of t h e observed crack growth law; thus, w e chose a value of
w a s within t h e " s t r i a t i o n " s tage which we were t r y i n g t o analyze.
Thus, questions a r i s i n g out of item ( 4 ) of t a b l e V should be
answered only i n r e l a t i o n t o the purpose intended t o be resolved, r a the r
than t o the academic question of the d e f i n i t i o n of a f a t igue crack.
It must be recal led, however, t h a t t he uses t o which t h e c u v e
lo t h a t
Item (5) of t a b l e V gets a t t he hea r t of t h e r e l a t i o n s sought fo r
predict ing both the e l a s t i c and p l a s t i c l i n e s of f igure 1 7 i n terms of
t h e d u c t i l i t y and t e n s i l e s t rength. By showing t h a t these two proper t ies
.
63
are predominantly e f f ec t ive i n governing the cyc l i c l i n e s of small smooth
specimens, we may conclude t h a t t h e important f a c t o r s governing the
i n i t i a t i o n of a crack and i t s very ea r ly s tages of propagation a r e t h e
manifestations of t h e capaci ty of a material t o deform.
d u c t i l i t y , which means high capaci ty f o r p l a s t i c deformation, and high
t e n s i l e s t rength , which i s a manifestation of t he capaci ty f o r e l a s t i c
deformation, a r e important cha rac t e r i s t i c s t h a t influence t h e a b i l i t y
of t he mater ia l t o resist crack i n i t i a t i o n and e a r l y s tages of crack
propagation.
Thus, high
Item ( 6 ) of t a b l e V r e f e r s t o the r e s u l t s of f i gu re 45 where, on
a l imi ted sca l e a t l e a s t , it i s seen t h a t a cumulative l i n e a r l i f e f r a c -
t i o n r u l e i s v a l i d when t h e major portion of t he l i f e i s involved i n
i n i t i a t i n g t h e crack.
high loca l i za t ion of s t r a i n , r e s idua l s t r e s s e s , e t c . , destroy t h e va l id-
i t y of the l i n e a r damage rule, but up t o the point where the crack i s of
s u f f i c i e n t s i z e t o introduce such complications, it would appear t h a t
The complexities associated with t h e crack i t s e l f ,
such a l i n e a r damage r u l e may be applicable.
those shown i n f igu re 45 a r e required f o r conclusiveness, and, of course,
we i m p l i c i t l y assume t h a t t h e mater ia l remains meta l lurg ica l ly s t ab le ,
and t h a t r a t e e f f e c t s (e.g. , creep) a r e absent.
Stage I1 Cracking
Table V I summarizes some important aspects of crack growth once it
O f course, more data than
has reached an appreciable length, and when the f a i l u r e propagation mode
i s one whereby progressive opening of t h e crack occurs normal t o the
d i r e c t i o n of loading. This period has been ca l led Stage I1 Cracking
by Forsyth ( r e f . 43). A s indicated by item (1) , the crack now moves
64
TABLE V I . - STAGE I1 CRACKING
(1) Crack penetrates gra in boundaries; assumes
c h a r a c t e r i s t i c t r a n s c r y s t a l l i n e na ture
( 2 ) Mechanism of crack growth may be r e l a t e d t o
b lunt ing and sharpening of crack under
tens ion and compression
(3) Independently formed cracks merge
( 4 ) Power laws of crack length and of p l a s t i c
s t r a i n descr ibe crack growth
(5) Not governed by l i n e a r l i f e f r a c t i o n r u l e
( 6 ) Residual s t r e s s induced by p r i o r s t r e s s
h i s t o r y i s important
e a s i l y across gra in boundaries and assumes the c h a r a c t e r i s t i c t r a n s -
c r y s t a l l i n e nature t y p i c a l of f a t igue failures.
govern the ex ten t of crack growth per cycle, and why i n f a c t it grows
only a ce r t a in d is tance i n one cycle, while extending f a r t h e r i n t h e
next cycle even though t h e nominal loading has not changed, have not
ye t been completely resolved. One explanation t h a t has been offered
i s based on the concept of b lunt ing and resharpening of t h e crack
( r e f . 44), as indicated i n item ( 2 ) of t a b l e V I .
grows during t h e t e n s i l e port ion of t h e cycle u n t i l t h e rad ius a t t h e
roo t of the crack increases (b lun t s ) enough t o r e l i e v e t h e s t r e s s con-
cent ra t ion s u f f i c i e n t l y so t h a t no f u r t h e r crack growth i s needed, a t
which time crack growth stops. The compressive port ion of t h e cycle
sharpens the crack, s o t h a t re -appl ica t ion of tension, f u r t h e r crack
growth i s needed t o r e l i e v e the s t r e s s concentrat ion again; however,
The f a c t o r s t h a t
Presumably t h e crack
65
t h i s concept provides only a partial explanation. O f g r ea t i n t e r e s t
would be an ana lys i s involving t h e res idua l s t r e s s e s because these
s t r e s s e s undoubtedly en te r i n t o the consideration of t h e amount of crack
extrusion, but a proper ana lys i s i s as ye t not ava i lab le .
Item (3) of t ab le V I r e l a t e s t o t h e c h a r a c t e r i s t i c of s tage I1
cracking involving the merging of independently formed cracks. Except
f o r cases involving very high cyc l ic l i v e s , f a t igue usua l ly involves
not one crack but many independently formed cracks t h a t merge, and crack
growth, i n p r t , cons is t s of t he merging of such cracks r a t h e r than j u s t
t he propagation of a s ingle crack. Most analyses of crack growth do
not , however, take such type of crack growth i n t o consideration.
A s indicated by item ( 4 ) of tab le V I , t he rate of crack length
growth during s tage I1 cracking i s governed by power l a w s of crack
length and p l a s t i c s t r a i n . Thus, when t h e crack i s small, t he growth
r a t e i s a lso small; as the crack length increases , i t s growth r a t e also
increases rapidly, s o t h a t i n t h e l a t t e r pa r t of t h e t es t t h e growth
r a t e may become ca tas t rophica l ly high compared with the r a t e i n t h e
e a r l y part of t he t e s t . Here a l so , i s a p a r t i a l explanation of why
notch s e n s i t i v i t y does not s ign i f i can t ly a f f e c t the f a t igue character-
i s t i c s i n t e s t s of t he type used t o e s t ab l i sh t h e l i f e r e l a t i o n s d i s -
cussed i n t h i s report . The l e s s notch sens i t i ve the mater ia l t h e
deeper t h e crack that can be withstood before f rac ture ; bu t the added
crack length would be achieved i n r e l a t i v e l y s o f e w cycles t h a t the
t o t a l number of cycles t o fa i lure i s not increased appreciably. It
must b e emphasized, however, tha t , while t h i s conclusion i s v a l i d f o r
t h e small, i n i t i a l l y unnotched specimens conventionally used i n low-
66
cycle-fatigue t e s t s , it may not be d i r e c t l y appl icable t o l a rge r s t ruc tu res ,
espec ia l ly i f they contain flaws o r notches,
Use of t he crack growth power laws does r e s u l t i n t h e p o s s i b i l i t y of
der iving a power l a w f o r cyc l i c l i f e i n terms of p l a s t i c s t r a i n (o r , i n
f ac t , i n terms of s t r e s s range) , bu t t h e implied exponents a r e not i n
complete agreement with the observed values. Par t of t h i s discrepancy
may be due t o t h e s impl i f ica t ions assumed i n t h e der ivat ion, bu t it may
also be t h a t t h e very process of studying crack growth r a t e s i n small
specimens, where the value of t he a c t u a l appl ied s t r a i n becomes be-
clouded once the crack i s la rge enough t o take meaningful measurements,
cons t i tu tes part of t h e d i f f i c u l t y . This problem requi res f u r t h e r
a t ten t ion .
Item (5) of t a b l e V I points out t h a t , when s tage I1 cracking i s
involved, a l i n e a r l i f e f r a c t i o n r u l e should not be expected t o apply.
The f a i l u r e of t he l i n e a r damage r u l e i n such appl ica t ions has not been
discussed here but i s wel l knwn from t h e l i t e r a t u r e . Thus, cases t h a t
involve an appreciable port ion of t he l i f e i n s tage I1 cracking - f o r
example, bending tests, or those of l a rge s t ruc tu res , e spec ia l ly i f
they contain notches or flaws - should not be expected t o follow a l i n e a r
damage rule. One of t h e most important reasons f o r t h e f a i l u r e of t he
r u l e (item ( 6 ) ) i s the p o s s i b i l i t y of introducing important r e s idua l
s t r e s s e s when cracks a r e present , whereas such s t r e s s e s a r e not con-
s idered i n the conventional appl ica t ion of t he theory. Other reasons,
however, also operate. One i s the na ture of t he crack growth curve, as
discussed i n connection with f igu re 48; another i s the l i f e value used
i n t h e denominator of t he cycle r a t i o s , as discussed i n connection with
67
f igu re 49. The subjec t of cumulative f a t igue damage of notched s t ruc tu res
i s a most important one, and i s receiving a g rea t dea l of a t t e n t i o n i n
many labora tor ies a t t h e present time.
F ina l Fracture
Although the f r a c t u r e process i s one of generat ing and propagating
a crack, the f a t igue l i f e i s usual ly taken as the number of cycles a t
which f r ac tu re occurs. F ina l f rac ture i s , however, a complex subjec t
i n i t s e l f . Most of t h e s tud ies cur ren t ly under way and the concepts
t h a t have been derived from them involve the assumption of l i n e a r elas-
t i c i t y i n the major sec t ion of the body that undergoes f r ac tu re . The
specimens from which our concepts of low and intermediate cycle f a t igue
a r e derived involve la rge p l a s t i c s t r a ins , usua l ly throughout t he f r a c -
t u r e cross sect ion. Thus, much needs t o be done t o br idge the gap be-
tween s tudies of f a t igue and of f rac ture , and hopefully t h i s subjec t
w i l l cons t i t u t e much of the content of those r epor t s i den t i f i ed i n
f igu re 1 by the dashed curve.
ACKNOWLEDGMENT
I n conclusion, I wish t o express deep g ra t i t ude t o my co-workers
whose ca re fu l experiments and helpful discussions have been of the
utmost value i n the preparat ion of t h i s r epor t : M. H. Hirschberg,
W. B. Bedesem, J. C. Freche, C. R. Ensign, and J. R. Kubancik.
68
REFERENCES
1. Thompson, N., and Wadsworth, N. J. : "Metal Fatigue", Advances i n
Physics, (Supplement t o Phi l . Mag.), Vol. 7, 1958, pp. 72-169.
2. Alden, T. H., "Basic Studies of Fqtigue Fracture i n Pure Metals",
Report No. 62-RL-2923M, General E l e c t r i c Research Laboratory,
February, 1962.
3. Avery, D. H., and Backofen, W. A. : "Nucleation and Growth of Fatigue
Cracks", i n Fracture of Sol ids , ed. by D. C. Drucker and J. J.
Gilman, John Wiley & Sons, Inc. , New York, 1963.
4. Grosskreutz, J. C . : "A C r i t i c a l Review of Micromechanisms i n
Fatigue", Proceedings of t h e Tenth Sagamore Army Materials Re-
search Conference, i n Fatigue - An I n t e r d i s c i p l i n a r y Approach,
ed. by J. J. Burke, N. L. Reed, and V. Weiss, Syracuse Universi ty
Press, 1964, pp. 27-59.
5. Gerberich, W. W . : "P l a s t i c S t r a ins and Energy Density i n Cracked
Plates. I. Experimental Techniques and Results", GALCITSM 63-23,
Graduate Aeronautical Laboratories, Cal i forn ia I n s t i t u t e of
Technology, Pasadena, Cal i forn ia , June, 1963.
6. Oppel, G. U., and Hill, P. W. : "St ra in Measurements a t t h e Root of
Cracks and Notches", Experimental Mechanics, Vole 4, No. 7,
Ju ly 1964, pp. 206-211.
7. Howie, A . , and Whelan, M. J. : "Diffract ion Contrast of Electron
Microscope
Royal Society of London, Vol. A263, 1961, pp. 217-237.
Images of Crys ta l La t t i ce Defects", Proceedings of t h e
69
8. Grosskreutz, J. C.: "Research on the Mechanisms of Fatigue",
Technical Documentary Report No. WADD-TR-60-313, Part 11,
Prepared under Contract No. AF 33(616)-7858 by Midwest Research
Institute, Dec. , 1963. 9. Bassett, G. A. , Menter, J. W. , and Pashley, D. W. : "Moird Patterns
on Electron Micrographs, and Their Application to the Study of
Dislocations in Metals", Proceedings of the Royal Society of
London, Vol. A246, 1958, pp. 345-368.
10. Lang, A. R.: "Studies of Individual Dislocations in Crystals by
X-ray Diffraction Microradiography", Journal of Applied Physics,
Vol. 30, 1959, pp. 1748-1755.
11. Dash, W. C.: !'The Observation of Dislocations in Silicon", Dis-
locations and Mechanical Properties of Crystals, John Wiley &
Sons, Inc., New York, 1957, pp. 57-67.
12. Hahn, G. T., and Rosenfield, A. R. t "Local Yielding and Extension
of a Crack under Plane Stress", Battelle kemorial Institute, 1964.
13. Wood, W. A., Cousland, S. McK., and Sargent, K. R. : "Systematic
Microstructural Changes Peculiar to Fatigue Deformation".
Metallurgica, Vol. 11, 1963, pp. 643-652.
Acta
14. Smith, R. W. , Hirschberg, M. H. , and Manson, S. S. : "Fatigue
Behavior of Materials Under Strain Cycling in Low and Inter-
mediate Life Range", NASA TN D-1574, April 1963.
15. Manson, S. S., and Hirschberg, M. H. : "Fatigue Behavior in Strain
Cycling in the Low and Intermediate Cycle Range", Fatigue - An
Interdisciplinary Approach, ed. by J. J. Burke, N. L. Reed, and
V. Weiss, Syracuse University Press, 1964, pp. 133-173.
7 0
16. Manson, S. S.: "Behavior of Materials Under Conditions of Thermal
Stress" Heat Transfer Symposium, University of Michigan, June 27-28,
1952, University of Michigan Press. Also:
Manson, S. S.: "Behavior of Materials Under Conditions of Thermal . Stress". NACA TN 2933, July 1953.
17. Manson, S. S.: "Thermal Stresses in Design, Part 19 - Cyclic Life of Ductile Materials", Machine Design, July 7, 1960, pp. 139-144.
18. Laird, C. , and Smith, G. C., "Initial Stages of Damage in High
Stress Fatigue in Some Pure Metals", Philosophical Magazine,
v01. 8, NOV. 1963, pp. 1945-1963.
19. Coffin, L. F., Jr. : "A Study of the Effects of Cyclic Thermal
Stresses on a Ductile Metal". Trans. ASME, Vol. 76, 1954,
pp. 931-950.
20. Weiss, V. : "Analysis of Crack Propagation in Strain-Cycling
Fatigue", in Fatigue - An Interdisciplinary Approach, ed. by J. J. Burke, N. L. Reed, and V. Weiss, Syracuse University
Press, 1964, pp. 179-186.
21. Palmgren, A. : "Die Lebensdauer von Kugellagern", ZVDI, Vol. 68,
No. 14, April 5, 1924, pp. 339-341.
22. Langer, B. F.: "Fatigue Failure from Stress Cycles of Varying
Amplitudes", Journal of Applied Mechanics, Vol. 4, No. 4,
December 1937, pp. A160-Al62.
23. Miner, M. A.: "Cumulative Damage in Fatigue", Journal of Applied
Mechanics, Vol. 12, 1945, pp. A159-Al64
7 1
24. Grover, H. J.: "An Observation Concerning t h e Cycle Ratio i n Cumu-
l a t i v e Ihmage", Symposium on Fatigue of A i rc ra f t S t ruc tures , ASTM,
STP No. 274, 1960.
25. Manson, S. S., Nachtigall , A. J., and Freche, J. C. : "A Proposed
New Relation for Cumulative Fatigue Damage i n Bending", Proceedings
ASTM, Val. 61, 1961, pp. 679-703.
26. Manson, S. S., Nachtigall , A. J , , Ensign, C. R., and Freche, J. C.:
"Further Inves t iga t ion of a Relation f o r Cumulative Damage i n
Bending". NASA Technical Memorandum TM X-52002, 1964.
27. Weibull, W. : "The Effec t of S ize and S t r e s s History on Fatigue
Crack I n i t i a t i o n and Propagation", Proceedings of t he Crack
Propagation Symposium, Cranfield, 1961, pp. 271-286.
28. Richart , I?. E., Jr., and Newmark, N. M.: "An Hypothesis f o r the
Determination of Cumulative Damage i n Fatigue", ASTM Proc.,
Val. 48, 1948, pp. 767-800.
29. Marco, S. M., and Starkey, W. L.: "A Concept of Fatigue Damage",
Trans. ASME, Vol. 76, 1954, pp. 627-632.
30. Corten, H. T., and Dolan, T. J. : "Cumulative Fatigue Damage",
Paper No. 2 of Session 3, In te rna t iona l Conference on Fatigue
of Metals, Vol. 1, Ins t . of Mechanical Engineers, London, 1956.
31. Freudenthal, A. M., and Hel ler , R. A. : "Accumulation of Fatigue
Damage", Fatigue of A i rc ra f t Structures , Academic Press, Inc. ,
New York, 1956, pp. 146-177.
72
32. Freudenthal, A. M., and Heller, R. A. : "On S t r e s s In t e rac t ion i n
Part I - 2024 Aluminum Fatigue and a Cumulative Damage Rule:
and SAE 4340 S t e e l Alloys", WADC TR 58-69, 1958 (AD 155687).
33. Henry, D. L.: "A Theory of Fatigue-Damage Accumulation i n S tee l .
Trans. ASME, Vol. 77, 1955, pp. 913-918.
34. Gatts, R. R.: "Application of a Cumulative Damage Concept t o
Fatigue", Trans, ASME, Ser ies D, Journal of Basic Engineering,
Val. 8 3 D , 1961 ( 4 ) , pp. 529-540-
35. Brown, G. W., and Work, C. E. : "An Evaluation of t he Influence of
Cyclic Pres t ress ing on Fatigue L i m i t " , Proceedings ASTM, Vol. 63,
1963, pp. 706-716,
36. Grover, H. J.: "Cumulative Damage Theories", Fatigue of A i r c r a f t
Structures , WADC Symposium, WADC, TR 59-507, August 11-13, 1959,
pp. 207-225.
37. Morrow, J., and Johnson, T. A. : "Correlation Between Cyclic S t r a i n
Range and Low-Cycle Fatigue Li fe of Metals", Materials Research &
Standards, Vol. 5, No. 1, January 1965, pp. 30-32.
38. Peterson, R. E. : "A Method of Estimating the Fatigue Strength of
a Member Having a Small E l l i p s o i d a l Cavity", In t e rna t iona l Con-
ference on Fatigue of Metals, In s t . of Mechanical Engineers,
London, 1956, pp. 110-117.
39. Ronay, M.: "On S t r a in Incompatibi l i ty and Grain Boundary Damage i n
Fatigue", Columbia University, I n s t i t u t e for t h e Study of Fatigue
and Re l i ab i l i t y , Technical Report No. 9, August 1964.
40. McEvily, A. J., Jr., and Boettner, R. C. : "A Note on Fatigue and
Microstructure", Fracture of Sol ids , John Wiley & Sons, New York,
1963, pp. 383-389.
,
73
41. Wood, W. A.: "Recent Observations on Fatigue Fa i lure i n Metals",
ASTM STP NO. 237, 1958, pp. 110-119.
42. Schijve, J.: "Analysis of the Fatigue Phenomenon i n Aluminum
Alloys 'I.
Amsterdam, NLR-TR-M2122, 1964.
National Aero- and Astronaut ical Research I n s t i t u t e ,
43. Forsyth, P. J. E.: "A Two-Stage Process of Fatigue-Crack Growth",
Proceedings of t he Crack Propagation Symposium, Cranfield, 1961,
pp. 76-94.
44. Laird, C., and Smith, G. C. : "Crack Propagation i n High S t r e s s
Fatigue", Philosophical Magazine, Vol. 7, 1962, pp. 847-857.
ANNUAL INCREASE
TOTAL PAPERS LISTED SINCE 1950 CS-34031
Figure 1. - Selected publications on fatigue (from ASTM references on fatigue).
Figure 2. - Factors associated with fatigue studies.
M
2024-0 ALUMINUM 6061-T6 ALUMINUM MILD STEEL
Figure 3. - Photoelastic coating method to determine strain distribution at notch tip.
CS-33791
(N. W . GERBERICH, 19631
MICROSCOPE OCULAR 8, CAMERA BACK
r SODIUM LIGHT
.- MICROSCOPE A,,, OBJECT IV E CLEAR, PLANE GLASS PLATE-._ 1 .,r SPECIMEN '. II , ,I' SURFACE
///////////////////////// SCHEMATIC OF OPTICS FOR OBSERVING
PHASE-INTERFERENCE FRINGES SPECIMEN UNDER
HIGH TENSILE LOAD cs-33790
(G. U. OPPEL 8 P. W. HILL, 1964)
Figure 4. - Phase interference method to determine strain distribution at notch tip.
I w
Figure 5. - Crack growth striations on fracture surface of U4- inch diameter polycarbonate res in specimen; 115 cycles to fracture.
,rREPLICATING ,‘ MATERIAL
I.
RACTURE SURFACE
APPLY SHADOW
\ ,”:P,”,”Z:N -REPLICATING
2. MATERIAL
3. REPLICATING MATERIAL I S DISSOLVED BY SOLVENT.
4. - REPLICA USED IN ELECTRON MICROSCOPE.
STRIATIONS ON FRACTURE SURFACE CHARACTERIZING FATIGUE CRACK GROWTH
MATERIAL: 7075-T6 ALLOY LIFE : 56,000 CYCLES
cs-34139
(BEDESEM 8 HARRELL - NASA) Figure 6. - Electron microscopy studies of fatigue fracture surface by replication. Material 7075-T6 alloy: life, 56 OOO cycles.
1 1 EI-&TRON INCREASED DIFFRACTION AT DISLOCATION--,.
SPECIMEN -1000% THICK
OBJECTIVE - I \ I ?zzi LENS
,-DIFFRACTED \ X #' BEAM OBJECTIVE APERTURE
rTRANSMlTTED BEAM
\ ,,- DISLOCATION IMAGE SCREEN
(HOWIE AND WHELAN, 1961) cs-33777
Figure 7. - Formation of dislocation image by transmission electron microscopy.
D I S L O C A T I O N S U B B O U N D A R I E S A F T E R 500 C Y C L E S
S U B G R A I N S A F T E R 5000 C Y C L E S
E L O N G A T E D S U B G R A I N S A F T E R 5000 C Y C L E S
Figure 8. - Use of transmlrrlon electron microscopy to observe formation of substructure i n aluminum. Total Strain range, 0.m4 life = 5x16 cycler
( G R O S S K R E U T Z , 1963)
ANALOG SHOWING APPEARANCE
PATTERNS OF DISLOCATIONS IN M O I R ~
(a) PARALLEL CASE (b) ROTATION CASE
EDGE DISLOCATION IN PALLADIUM SINGLE CRYSTAL GROWN ON GOLD SINGLE CRYSTAL
cs-33787
(G. A. Bassett e t al . , 1958)
Figure 9. - M o i r i method used to observe dislocations in crystal lattice by electron microscopy.
(3-33785
(A. R. LANG, 1959)
Figure 10. - Observation of dislocations in si l icon by X-ray diffraction.
Figure 11. - Observation of dislocations by decoration. Frank-Read source in silicon observed using copper decora- tion. CS-33788
( W . C. DASH, 1957)
N 0 TCH-,
SURFACE
NOTCH-.,
MIDSECTION CS-33786
(HA" 8 ROSENFELD, 1964) Figure 12. - Etching techniques to reveal plastic
zone.
SURFACE :SECTIONING PLANE
SPECIMEN: - -- - -
TAPER MAGNIFICATION, COSECANT 8 NOTCH DEPTH ON SECTION
TAPER SECTION OF BRASS NORMAL, A TAPER, B SPECIMEN IN TORSION
(533789
Figure 13. -Observation of fatigue damage at specimen surface by taper sectioning.
(W A WOOD,et.ol., 19631
R O O T 0.000 7"
MAX. CYCL I c
S T R E S S , KS I
4 0
30
2 0
I O
S - N CURVES FOR NOTCHED 2 0 1 4 - T 6
0 6 0" SHE E T
r F l R S T O B S E R V A T I O N OF /I CRACKS (.0025")
i 4 R = . 1 4 2 8
-4 \
IO' I o3 io5 I 07 CYCLES
CS-33765
K L I M A , L E S C O , F R E C H E , 1964
Figure 14. - Ultrasonic technique for detecting early fatigue cracks.
280 t STRESS 240 RANGE,
200
I 6 O t
ELASTIC r r MODULUS
I
CYCLES
CS-22779
I I
I20
0 .01 .02 .03 .04 3/4 IO IO2 IO3 IO4 0 300 600 900 1200 1500 STRAIN RANGE
la1 Static and cyclic stress-strain character- Ibl Stress range as lunction of applied cycles; Id Stress range as function of applied cycles: linear scale. istics. lqar i thmic scale.
440 -
400 -
360 -
320 - - o) x
J a 4 280-
Y 240- II:
o)
LL I- o)
200 -
160 -
ELASTIC MODULUS LINE
I
Figure 15. - Static and cyclic stress-strain characteristics of 304 stainless steel.
I
o .OI .02 03 .04 .05 314 IO 10' 103 lo4 o IO00 2000 3000 Strain range Cycles
la) Static and cyclic stress-strain charac- Ibl Stress range as function of applied cycles: I C 1 Stress range as function of applied cycles: linear
CS-22781 teristics. lqar i thmic scale. scale.
Figure 16. - Static and cyclic stress-strain characteristics of 4340 steel.
M
I
.IO RANGE STRAIN
.o I
.oo I
STRAIN RANGE
I I I I IO-' loo 10' lo2 lo3 lo4 lo5 lo6 10'
CYCLES TO FAILURE CS-22507
Figure 17. - Total strain as combination of two linear relations.
- I IO' io4 105
CYCLES TO FAILURE 4
Figure 18. - Method of predicting axial fatigue life based on ductility, fracture strength, and ultimate strength.
FRACTURE STRESS,
KSI
( I+D)uu ,KSI (333668
Figure 19. - Fracture stress against function of t rue ductility and ultimate strength.
--- PI ELASTIC LINE AT 1/4 CYCLE
ELASTIC LINE AT Id CYCLES,
P2
.om
.01m
.m80 .+
.mul
. 0020 0
PLASTIC LINE AT 104 CYCLES
r ELASTIC
LIFE, CYCLES
0 20 40 60 80 100 REDUCTION I N AREA, 'b
I I 1 1 I I 1 I I l l I I l l I I I 1 I .02 . 1' . a . 3 . 4 .6
PLASTIC LINE AT 10 CYCLES, P3 cs-33782
Figure 20. - Predicting axial fatigue from tensi le data by four-point correlatlon method.
.4
.2
w
- -
- I I I
- 29 MATERIALS
- - -
*€PI
D a6 -
.01-- - -
- -
Life, cycles cs-33674 Figure 21. - Relation between plastic strain ductility and cycles to failure.
LIFE, CYCLES C S -3420
Figure 23. -Model for method of universal slopes.
REDUCTION IN AREA, %
Figure 24. - Predicting axial fatigue from tensile data
4 c m c CL W V L
Q) +- c I .-
cs-33779
by method of universal slopes.
.I 0
.IO
. I O
TOTAL STRAIN RANGE ,lo
,001 - 100 i o 2 104 i o 6
TOTAL STRAIN RANGE
.I 0 c 00 ARD
.IO L<ut:iET 0 EXPERIMENTAL
.I 0 _ - _ PREDICTED BY FOUR- 1 % POINT CORRELATION
PREDICTED BY 300 __ UNIVERSAL S L O P E S .IO k M
.o . l o b I
-k ANNEALED
001 - 100 102 lo4 lo6
CYCLIC LIFE cs-33669
Figure 25. - Comparison of predicted and experimental axial fatigue life for lw-a l loy and high-strength steels.
r r
.I 0
.IO
. IO
.IO
.IO
.o I
0 EXPERIMENTAL PREDICTED BY FOUR-
POINT CORRELATION
UNIVERSAL SLOPES
.I 0 -PREDICTED BY
.IO
.01
.oo I loo io2 104 106
cs-33667 CYCLIC LIFE Figure 26. - Comparison of predicted and experimental axial fatigue life for stainless steels and high-temperature alloys.
r r
TOTAL STRAIN RANGE
0 EXPERIMENTAL
_- - PREDICTED BY FOUR-
- PREDICTED BY
POINT CORRELATION
UNIVERSAL SLOPES
(3-33666
Figure 27. - Comparison of predicted and experimental axial fatigue life for nonferrous metals.
CYCLIC L IFE
I- / 1 29 MATERIALS . /
' . . . . . /.: .:. . . . . . . . . . . . . . . ,
MEASURED LIFE,
CYCLES
L ... ..
I o4 . . .
I .5 2 5 I O 2 0
35 6 2 09 9 4 97
CS-33672 PREDICTED LIFE, CYCLES
Figure 28. - Measured as function of predicted life determined by four-point correlation
PREDICTED TOTAL STRAIN
RANGE
. : / 29 MATERIALS
1.or 29 MATERIALS
.I F
.OOl .01 .I I .o MEASURED TOTAL STRAIN RANGE cs-33673
Figure 29. - Predicted as function of measured total strain range by four- point correlation.
%, STRAIN OF FACTOR DATA
1 . 1 33 I . 2 60 I .5 90 2 96 5 100
MEASURED LIFE,
CYCLES
-00 I I I 1 1 1 1 1 1 1 I I I I I l l l l .OOl .01 .I I .o
MEASURED TOTAL STRAIN RANGE cs-33673
Figure 29. - Predicted as function of measured total strain range by four- point correlation.
1 0 ~ ~
I- / 29 MATERIALS
, " . . ,/ . . .. . . .. .. . ,
1o4k
/ ... . . :. : :....,. . . . . .. . . ' : .
. . . . . . / ,. ._ ._ . :'< .
%, LIFE OF
FACTOR DATA I .5 37 2 59 5 90
I O 96 20 99.5
PREDICTED LIFE, CYCLES CS-33671
Figure 30. - Measured as function of predicted life determined by method of universal slopes.
ls0F 29 MATERIALS /
TOTAL STRAIN RANGE, A6
STRAIN FACTOR
1.1 1.2 1.5 2 5
%, OF
DATA 31
5 5 84 96
100
MEASURED TOTAL STRAIN RANGE CS-33670
Figure 31. - Predicted as function of measured total strain range by method of universal slopes.
N' = A(AC-AC~ P EXPE R I ME N T A L METHOD OF UNIVERSAL SLOPES \ * POINTS FITTED TO GET
I
--- lr
.IO0
.OlO
.OOl
CONSTANTS A,A(o AND v
I I '\ I
I I I I I I I I I I I I
I T-* & I
1 I I I I I 10' lo2 lo3 lo4 lo5 lo6 IO?
CYCLIC LIFE, Nf CS-34025
Figure 32. - Alternate equation for life in terms of total strain range.
TOTAL STRAIN RANGE
.IO
.IO
.IO
.IO
.IO
.o I
FIT OF THREE POINTS ON- EXPERIMENTAL CURVE CURVE PREDICTED BY
U N I V E R S A L SLOPES
VASCOJET
X-HARD
.I 0
CVM .IO
HARD
ANNEALED
,001 - IOO lo2 lo4 lo6
.0011 I I I loo IO* lo4 lo6 c5-311337
CYCLIC LIFE
Figure 33. - Comparison of predicted and experimental axial fatigue life for Iw-alloy and hlgh- strength steels using alternate equation involving endurance l imi t
t . FIT OF THREE POINTS ON- EXPERIMENTAL CURVE CURVE PREDICTED BY UNIVERSAL SLOPES
TOTAL STRAIN RANGE
.IO
.IO
.IO
.o I
.IO
.IO
.IO
.o I TAINLESS STEEL
? ,
TOTAL STRAIN RANGE
FIT OF THREE POINTS ON- - EXPERIMENTAL CURVE --- CURVE PREDICTED BY N,= A ( A ~ - A ~ , S /
UNIVERSAL SLOPES
MAGNESIUM
io
.IO
.IO
,100
100 102 104 106
LIFE
Flgure 35. - Comparison of predicted and experimental axial fatigue llfe for nonferrous metals using alternate equation involving endurance limit
50
40
35
220
I I 45[ I-\ RC
HARDNESS,
Rc
I
0 NASA DATA STRENGTH, TENSILE 2 0 0 1 w- VENDOR DATA
I I
KSI 180
160 I I I Q U I I
REDUCTION IN AREA, PERCENT
IZOD IMPACT,
FT-LB
40 6, tK;RA 20 IZOD
TEMPERING TEMPERATURE, "F
0 400 600 800 lo00
CS-33676
Figure 3 6 - Mechanical properties of type 410 stainless steel quenched f rom 1850" F.
.
. O O o l i
STRAIN RANGE
A 0 850" F I I
TEMPER A 0 400" F
. Gu1
- / I / I / I
1 / I
0
0
AcD = 0.080, Nf = 115
AcD = 0.032, Nf = 1300 .24
0 Nf = 115
. *I 01
0 Nf 13CfJ
Fatigue life, percent
(a) Rectangular coordinates. (b) Semilogarithmic coordinates.
Figure 38. - Crack g r w t h curves for polycarbonate resin.
PERCENT OF LIFE TO
100
60
-1/3 Nf
6 410 STEEL 400°F TEMPER 0 410 STEEL 85OOF TEMPER A 2 0 2 4 - T 4 ALUMINUM
4130 HARD STEEL 0 POLYCARBONATE 0 PURE AI AND PURE Ni
0
CRACK INITIATION 40-
0
20 -
(FROM LAIRD AND SMITH, 1963) 0 I I IO' 162 103 lo4 lo5 io6 cs-34033
Cyclic life, N f
Figure 39. - Relation between percent of life to crack initiation and fatigue life.
PLASTIC STRAIN RANGE
LIFE BASEDON FINAL FRACTURE
SLOPE = -0.5 CRACK INITIATION
POINTS COMPUTED FROM CRACK INITIATION CURVE, FIGURE 39
.0051 I IO0 IO' IO2 lo3 lo4
cs-34030 LIFE, CYCLES Figure 40. - Relation between plastic strain range and life as determined by
crack initiation and by fracture.
1. 4 r
LOAD RANGE,
LB
1. -- c m
m : s.1.0 n - 0 . 2 q - 1.0
exponent, w-0.56
4
. 2 0 . 2 . 4 . b . a 1 . 0
PLASTIC FRACTURE PARAMETER, nq
Figure 41. - Relation between plastic strain exponent and other constants entering into crack growth and failure laws.
CS-35002
- SINGLE STRAIN LEVEL TESTS HIGH STRkiN FiRST
----- LOW STRAIN FIRST
---
p 0 2 4 - T 4 ALUMINUM
A~s.070 r l ~
8000- B G'
Nf = 657
6000
5000 100 1000 I 10
C I 0 20 40 60 80 100 120
CYCLES I LIFE (a) Load range as function of num- (b) Load range as function of percent of life.
ber of cycles applied. CS-34032 Figure 42. - Cyclic strain hardening under two level fatigue tests,
c
SINGLE STRAIN LEVEL TESTS
9000
8000 LOAD RANGE,
700C
600C
500(
--- HIGH STRAIN FIRST ---___ L O W STRAIN FIRST
2024-T4 ALUMINUM r
0 50 100 0 50 100 % LIFE
(a) Single change of strain level at 25 percent of life.
(b) Single change of strain level at 50 percent of life.
M I ro (D w P
0 50 100
(c) Strain levels changed at 25
cs-34034
and 50 percent of life.
Figure 43. - Cyclic strain hardening under multilevel fatigue tests.
- SINGLE STRAIN LEVEL TESTS --- HIGH STRAIN FIRST ----- LOW STRAIN FIRST
20 000
16000 RANGE,
LB D
12 000 0 50 IO0 ’ 0 50 100
% LIFE
r
k .----- I C S-34022 u 0 50 100
(a) Single change of s t ra in level at 25 percent of life.
(b) Single change of s t ra in level a t 50 percent of life.
(c) S t ra in levels changed a t 25 and 50 percent of life.
F igure 44. - Cyclic s t ra in softening under mult i level fatigue tests.
2024-T4 L29 L28 L24 L 33 L34
ALUMINUM
rl M cn cu I w
-- - * 1-
T33 T34
T38
T40
Ti -6AI-4V
SPECIMEN I
I
T 4 5 1 -
T13-- -
RELATIVE CRACK
LENGTH
0 20 40 60 80 100 % OF LIFE CS-34017
Figure 45. - Cumulative fatigue damage obtained by strain cycling of smoot h specimens.
CYCLES
(a)
Figure 46. - Idealized crack growth
% FATIGUE LIFE ~5-34026
(b)
behavior satisfying linear damage rule.
RELATIVE CRACK LENGTH
RELATIVE CRACK
LENGTH
6r IC' 2-
0 200 400 600 800 1000 CYCLES
0 20 40 60 80 100 % FATIGUE LIFE
(a) (b) CS-34027
Figure 47. - Idealized crack growth behavior showing need for deviation from linear damage rule.
ORDER OF 9+!!2 1 LOAD I Nl N2 1'; 1 HIGH- LOW .50+( P.75) LOW- HIGH .75 + ( I -.SO)
0 20 40 60 80 100 CYCLES % FATIGUE LIFE
(a) (b) CS-34019
Figure 48. - Idealized crack growth behavior complying with experimental trends.
,
rl M cn cu w
RELATIVE CRACK LENGTH
"REMAINING LIFE-'
0 200 400 600 800 1000 CYCLES
C'
: 0 20 40 60 80 100
%REMAINING LIFE (a) Relative crack length as func t i on of num- (b) Relative crack length as func t i on of per-
ber of cycles applied. cent rema in ing life. CS-3423
Figure 49. - Idealized crack growth behavior fo r precracked specimen.
STRESS AMPLITUDE I
40
KSI
IO' a
I o2 I o3 io4 I os I o6 CYCLES TO FAILURE CS-27138
Figure 50. - Fatigue data under axial loading and in rotating bending for 4130 steel.
0 .O IO5 H
(a) Strain distribution for life (b) Stress distribution from (c) Contribution of strip to of 1OOO cycles. (a) and cyclic stress-strain bending moment.
CS-27136 curve. . .". .L.. ,
Figure 51. - Determination of bending moment supported by circular section for selected value of surface
strain. M = 2JR dM = 4 i R { D ox dx = 113 OOO in. -Ib.
300
I 5 0
100
50
PREDICTION 0 E X P E R I M E N T A L F-
CONVENTIONAL STRESS
KSI
4130 (HARD) STEEL 4130 (SOFT) STEEL
60
40
20
101 lo2 103 104 lo5 lo6 10' io2 io3 io4 io5 io6 FATIGUE L I F E , CYCLES cs-34029
7 0 7 5 - T 6 ALUMINUM 2 0 2 4 - T 4 ALUMINUM
Figure 52. - Prediction of flexural fatigue from reversed axial strain cycling behavior.
CYCLIC LIFE
(a) Total s t ra in range l i fe relat ion for material.
A E
IC) Cyclic stress range - st ra in range relat ion for material.
ROT
a 0
(b) St ra in distr ibution in cross section of rotat ing beam.
(d) Hysteresis loop for points at radius r of rotat ing beam.
CS-33783
Figure 53. - Determination of stress distr ibution in a rotat ing beam f rom axial strain cycl ing fatigue data.
:.HYSTERESIS 0 =TAN- 1 % (%) ANGLE
R 4Mxx NOMINAL OMAX. M,, i -
7A3 Y
CS-33778 \
Figure 54. - Determination of bending moment a n d hysteresis angle in a rotat ing beam.
300- - PREDICT ION 300
250- 00 EXPERIMENTAL 250-
200
I50
100 100
5 0 (3-
O I I I I I I I I 0
CONVENTIONAL :::; 5 0 ".-, STRESS
-
Rl
- -
- -
STRAIN RANGE -- 1% A T io4 CYCLES
2% AT io3 CYCLES -----
120- 2 5 0
200
loo! 60 80 41".., f 40- 20- 0-0- 5 0 -
O I I I I I I I I 0
M I
-
- \ 0- D-
1 1 1 1 1 1 1
\ ' A
I 4
I B
t C D
cs-34080 (WOOD, 1958)
Figure 57. - Schematic view of crack nucleation.
(SCHIJVE, 19641
Figure 58. - Crack propagation along slip lines in aluminum.